The Leptosphaeria maculans – Leptosphaeria biglobosa species complex in the American continent




Stem canker of oilseed rape (canola, Brassica napus) is associated with a species complex of two closely related fungal species, Leptosphaeria maculans and L. biglobosa. Of these, L. maculans is the most damaging and develops gene-for-gene relationships with the host. Here, a wide scale analysis of the L. maculans - L. biglobosa species complex was performed throughout the American continent (23 locations from Chile to Canada) plus several locations in Western Australia for comparison purposes, based on a collection of 1132 isolates from infected tissues of a susceptible cultivar. Fungal species were discriminated on the basis of morphological, phytopathological and molecular criteria and showed that L. biglobosa was closely associated with L. maculans in most of the locations. Multiple gene phylogeny using sequences of ITS, actin and β-tubulin confirmed the prevalence of the L. biglobosa‘canadensis’ sub-clade in Canada, whereas up to three different sub-clades of L. biglobosa were found in Georgia (USA). Race structure of L. maculans was investigated using a combination of pathogenicity tests and PCR amplification of avirulence alleles AvrLm1, AvrLm4 and AvrLm6. Three contrasting situations were observed: (i) race structure in Ontario, Chile and Georgia was related to that of European and Western Australian populations, with a low race diversity; (ii) only one race was found in Mexico, and not found outside of this country; (iii) a large diversity of races was observed in central Canada (Manitoba, Alberta and Saskatchewan) with very specific features including maintenance of avirulence alleles absent from Europe, absence of the AvrLm7 allele common in Europe (or eastern Canada) and wide location-to-location variability.


Phoma stem canker (blackleg), caused by the fungus Leptosphaeria maculans, is the most economically important disease of oilseed rape (Brassica napus) worldwide (Howlett, 2004; Fitt et al., 2006). In this plant-pathogen system, gene-for-gene interactions commonly occur and major resistance genes (Rlm genes) are widely used to control the disease (Delourme et al., 2006). Nine of these (Rlm1–Rlm9) have been characterized following the genetic identification of their counterpart avirulence genes (AvrLm1–AvrLm9) (Rouxel & Balesdent, 2005) and have proved to be extremely efficient to control the pathogen whenever they are faced with avirulent populations of the fungus (Ansan-Melayah et al., 1997). Unfortunately, being highly efficient, they are often grown on large acreages and thus exert a strong selection pressure on the fungal populations (Rouxel et al., 2003a). Leptosphaeria maculans is a pathogen with a great ability to adapt to and overcome such novel host resistance as it combines a mixed reproduction system with a prevalence of sexual reproduction, very large population sizes and long-range dispersal of ascospores (Rouxel & Balesdent, 2005; West & Fitt, 2005). In this respect, the literature shows that it can adapt very rapidly to novel selection pressures such as the use of new resistance sources in oilseed rape, either in experimental systems where artificial inoculum was increased (Brun et al., 2000) or in actual agricultural systems following the commercial success of a cultivar with a novel resistance source (Li et al., 2003; Rouxel et al., 2003a; Sprague et al., 2006). In all cases, only three to four years were sufficient for the pathogen to be able to overcome the novel resistance source leading to inability of the resistance to control the disease. It is thus crucial to have an extensive knowledge of populations of the pathogen present at a given time, including site-to-site variations, and to know their race structure (occurrence of avirulence (Avr) alleles within the population and the combination of Avr alleles present in isolates) before using any novel resistance source. Knowledge (and dynamic survey) of race population structure is thus an essential tool, both in choosing the best-adapted resistance sources to be used locally and also for proposing strategies for durable use of such resistance sources.

Whereas such extensive data has been gathered throughout Europe (Balesdent et al., 2006; Stachowiak et al., 2006), and to a lesser extent in Australia (Balesdent et al., 2005; this study), only very limited analysis of race structure of L. maculans populations has been performed in America. This is surprising since Canada is one of the main B. napus-growing countries in the world and the first dramatic epidemic of L. maculans was reported on cabbage in Wisconsin (Henderson, 1918) suggesting virulent isolates of L. maculans were present at this time in the USA.

In this pathosystem, there is an additional level of complication in that the disease is associated with a complex of species, L. maculans and L. biglobosa, two closely related species, and at least eight subclades (Mendes-Pereira et al., 2003; Voigt et al., 2005). Resistance genes towards L. maculans do not affect L. biglobosa and little is known about resistance to L. biglobosa (Fitt et al., 2006). These two species are very similar, but numerous methods to discriminate them were reported until the advent of sequence-based molecular tools (Mendes-Pereira et al., 2003). Even though they share similar epidemiology, identical infection strategies, and related ecological niches (Fitt et al., 2006), and are often found together in the tissues of individual infected plants, the symptoms and severity of disease they cause are often very different (West et al., 2002). However, at the leaf level, symptoms of L. biglobosa can often be confused with those caused by pathogenic Alternaria spp. of oilseed rape, or with those of cultivars expressing moderate resistance to populations of L. maculans. Following systemic colonization of plant tissues, L. maculans isolates cause damaging basal stem canker (crown canker), whereas L. biglobosa isolates cause pale brown lesions with a dark margin on the upper stem (termed phoma or upper stem lesions). Leptosphaeria maculans isolates have been shown to be highly specialized pathogens, developing gene-for-gene interactions with all their Brassica hosts (B. napus, B. rapa, B. oleracea, B. juncea, etc.). By contrast, such specialized interactions have not been observed with L. biglobosa isolates (Vincenot et al., 2008). Another dissimilarity between L. maculans and L. biglobosa regards their geographic distribution. Whereas both species are widely distributed worldwide, there are indications that L. maculans is a currently expanding species that historically colonized countries where L. biglobosa was prevalent, such as Poland and central Canada (Fitt et al., 2008). Molecular phylogeny studies support the view that L. maculans is a younger expanding species and reveal that L. maculans is a monomorphic species whereas the L. biglobosa species encompasses six distinct sub-clades (Mendes-Pereira et al., 2003; Vincenot et al., 2008). The most common sub-species L. biglobosa‘brassicae’ is found in most oilseed rape growing regions of the world except central Canada and Australia and is closely associated with L. maculans on oilseed rape plants (Mendes-Pereira et al., 2003; Fitt et al., 2006). Finally, within the extremely diverse L. biglobosa sub-species, recent work strongly suggests differences in ability to infect oilseed rape. Some sub-species such as L. biglobosa‘brassicae’ and, to a lesser extent, L. biglobosa‘occiaustralensis’ show high levels of aggressiveness towards B. napus and some evidence of adaptation to specific cultivars (cvs) (Vincenot et al., 2008).

In this study, a wide-scale sampling and characterization of the L. maculans – L. biglobosa species complex in America is described. Fields with trap plants harbouring no Rlm genes were set up in 2004–2005 and 2005–2006 in different locations in Chile, Mexico, Georgia (USA) and Canada, with a special emphasis on Canada, with four provinces and 19 locations investigated. Isolates were obtained from infected stem and root tissues and characterized at the species/subclade level using multiple-gene genealogies, and at the race level for L. maculans isolates, using a combination of pathogenicity tests on differential genotypes and PCR-amplification of three avirulence genes. The results obtained were compared with European and Australian populations.

Materials and methods


The isolates obtained in this study originated from Mexico (Zacatecas, 28 isolates, 2005; Gout et al., 2007), Georgia (Plains and Griffin, 121 isolates, 2005), Chile (Temuco, 130 isolates, 2005) and 19 different sites in four provinces of Canada (Alberta, Ontario, Manitoba and Saskatchewan, 764 isolates, 2005 and 2006) (see Table 1 for a listing of L. maculans isolates listed here). The samples were obtained from a trap cultivar (Westar in most cases; broccoli cultivars harbouring none of the currently known Rlm genes for Mexico) planted according to the protocol established by Balesdent et al. (2006). However, in Chile 24 isolates were obtained from cv. Surpass400 harbouring Rlm1 (Van de Wouw et al., 2009) and 68 others were from undetermined spring cultivars postulated to be devoid of any R genes in view of the race structure on these cvs being similar to that obtained on Westar (see Results section). One thousand and forty three single hyphal tips isolates were obtained from roots or stems of infected residues and hyphal tip purified as described by West et al. (2002). For comparison purposes, 89 single-conidium isolates from Western Australia were also analyzed (Vincenot et al., 2008). These were obtained from leaves of cv. Westar in 2004 and in seven locations in Western Australia. Finally, data for EU isolates described in Balesdent et al. (2006) and Stachowiak et al. (2006) were also included in population analyses. All fungal cultures were maintained on V8 juice agar and highly sporulating cultures were obtained according to the procedure described by Ansan-Melayah et al. (1995). For pathogenicity tests and DNA extraction, conidia were collected from 12–15-day-old sporulating cultures (Balesdent et al., 1998).

Table 1.   Characteristics of the race structure of American Leptosphaeria maculans populations
Site Region/ province/ countryNo. of
Margalef Indexa Simpson Indexa No. of racesaNo. of virulence
per isolatea
  1. aRaces as defined by the combination of avirulence alleles at loci AvrLm1-AvrLm4, AvrLm6, AvrLm7 and AvrLm9.

  2. bIsolates obtained from the resistant Surpass 400 cv. in Chile are not included in the Table.

Elm CreekManitoba253·570·90253·32
Birch HillsSaskatchewan395·660·987103·41

Pathogenicity tests

Races of L. maculans were determined by inoculating all of the isolates onto a B. napus differential set, comprised of fixed lines or commercial cultivars with as few Rlm alleles as possible (Table 2). Depending on the Avr allele complement of the isolate, different strategies, including the use of better-adapted plant genotypes and PCR amplification of Avr genes were used in race determination (see Results).

Table 2.   List and Rlm gene complements of Brassica napus cultivars used to characterize the collection of isolates of Leptosphaeria maculans
Cultivar/lineResistance gene present
Jet neuf and PixelRlm4
04-35-2-5 and 00-136-2-1Rlm1,Rlm7

Isolates were inoculated onto cotyledons according to established protocols (Balesdent et al., 2006). Symptoms were scored on 10–12 plants at 15 and 21 days after inoculation, by using the IMASCORE rating scale comprised of six infection classes (1 to 6). Classes 1 to 3 correspond to avirulent isolates (AvrLm) and classes 4 to 6 correspond to virulent isolates (avrLm) (Balesdent et al., 2005). Seven of the nine Avr alleles currently known were characterized in this study, due to insufficient availability and homogeneity of plant material harbouring Rlm5 (B. juncea genotypes) and Rlm8 (B. rapa genotypes). Race terminology was as established by Balesdent et al. (2005) with each L. maculans race being identified by the successive numbers of Avr loci for which the isolate has been characterized and is avirulent, preceded by the letters ‘Av’, to show that only Avr alleles are indicated. The Avr loci for which the genotyping has not been possible (here mostly AvrLm5 and AvrLm8) are indicated in parentheses.

DNA extraction, PCR and sequencing

Genomic DNA was extracted from conidial suspensions using the DNeasy 96 Plant Kit and the QIAGEN BioRobot 3000 in accordance with the manufacturer’s recommendations.

A multiplex PCR, developed to rapidly characterize the mating type of L. maculans isolates (Cozijnsen & Howlett, 2003), was used to check that isolates belonged to the L. maculans species. PCR amplification of the ITS region using primers PN3 (5’-CCGTTGGTGAACCAGCGGAGGGATC) and PN10 (5’TCCGCTTATTGATATGCTTAAG) was done according to Mendes-Pereira et al. (2003). Two different couples of AvrLm1 primers were used (internal primers AvrLm1 IntU: TAGCTCCCCAGCTACCAAGA and AvrLm1 IntF: ACGTTGTAATGAGCGGAACC, external primers 133D2-AVR1U: CTATTTAGGCTAAGCGTATTCATAAG and 133D2AVR1L: GCGCTGTAGGCTTCATTGTAC) (Gout et al., 2006). One pair of external primers located in the promoter and in the 3’UTR of AvrLm6 was used to amplify AvrLm6: AvrLm6ext-U, TCAATTTGTCTGTTCAAGTTATGGA and AvrLm6ext-L, CCAGTTTTGAACCGTAGTGGTAGCA (Fudal et al., 2009). Finally, one pair of primers AVR47ext-Up (TATCGCATACCAAACATTAGGC), located in the promoter of the gene, and AVR47ext-Lo (TTGTTAGCGGTTGATCCATC), located in the 3’ UTR of the gene were used to amplify AvrLm4 (Parlange et al., 2009).

PCR was performed in a 9700 Applied Biosystems thermal cycler (Applied Biosystems) or an Eppendorf Mastercycler EP Gradient thermocycler (Eppendorf) as described previously (Gout et al., 2006; Fudal et al., 2007). For the sequencing of ITS, actin and β-tubulin, PCR was performed according to Mendes-Pereira et al. (2003) or to Voigt et al. (2005). Sequencing of PCR fragments using the Beckman Coulter CEQ 8000 automated sequencer (Beckman Coulter) was done in accordance with the supplier’s recommendations.

Data analysis

Populations (isolates from a given site, or a given country/province) were compared for frequency of each resolved avirulent allele and for race complexity of the isolates (number of virulences per isolate). In addition, the Margalef index, which measures the richness in species (here, in races) of a population, and the Simpson index of richness were used to analyze population diversity (Pinon & Frey, 1997). Whenever the identification of the allele present at one Avr locus was possible, the data was recorded as 0 (virulent allele) or 1 (avirulent allele). The data matrix obtained for all field isolates and Avr loci was used to calculate genetic distances (Nei, 1973) between genotypes. The resulting matrix was used for a neighbour-joining cluster analysis using the ‘Populations’ software (Langella, 1997). Trees were edited by using the TreeView and TreeDyn software (Chevenet et al., 2006; Page, 1996).

Phylogenetic analysis of L. biglobosa isolates

Preliminary sequence alignments were performed using ClustalX (Thompson et al., 1994) and included published ITS, actin and β-tubulin sequences from L. biglobosa isolates of the species complex (Mendes-Pereira et al., 2003; Voigt et al., 2005). Alignments were checked manually for ambiguities and adjusted when necessary using Gene Doc 2·6·002 (Nicholas et al., 1997). Following manual editing, genetic distances were calculated and Neighbour-Joining (NJ) trees were prepared using the Draw N-J tree option of ClustalX. Parsimony analyses were performed using the Phylogenetic Analysis Using Parsimony package (paup* 4·0b10; Sinauer Associates; Swofford, 1998).


Presence and characterization of L. biglobosa in the American continent

Of 1043 American isolates obtained here, 85 failed to amplify the Mating Type locus using L. maculans-specific MAT primers. These 85 isolates also showed morphological traits, phytopathological features and ITS-PCR product size compliant with those of L. biglobosa (data not shown). Leptosphaeria biglobosa isolates were found in all sampled countries/provinces, except Ontario and Mexico (Fig. 1). In Chile, they were only isolated from a L. maculans-resistant cv. (Surpass 400), as already observed by Vincenot et al. (2008) for Western Australia populations. In all other countries/provinces L. biglobosa isolates could be obtained from the susceptible trap cultivar, Westar at a weak but steady rate within the samples. The main exception was Alberta where more L. biglobosa isolates than L. maculans isolates were obtained from Westar (57·1–64·1% of the samples depending on the locations; Fig. 1 and data not shown). Finally, and except for Alberta, site-to-site variations within a single country were weak (0–10·7% of the samples; data not shown).

Figure 1.

 Ratio (%) of Leptosphaeria biglobosa (white bars) and L. maculans (black bars) isolates obtained from different locations within the USA, Chile, Mexico and Canada. Eighty-five L. biglobosa isolates out of a total of 1043 American isolates were obtained.

Multiple gene phylogeny of American L. biglobosa isolates

A total of 46 ITS sequences were obtained for central Canada L. biglobosa isolates. Due to complete identity between these sequences (data not shown), only 14 of these were included in the multiple gene phylogeny analysis. In total, 23 actin and β-tubulin sequences for USA, Chile and Canada L. biglobosa isolates were generated and submitted for phylogenetic analyses by comparison with the previously published sequences from other members of the species complex (Mendes-Pereira et al., 2003; Voigt et al., 2005; Vincenot et al., 2008). The ITS alignment resulted in a dataset of 502 characters with 29 (5·8%) parsimony informative characters. Alignment of the actin sequences resulted in a dataset of 838 characters of which 39 (4·6%) were parsimony-informative. Alignment of the β-tubulin sequences resulted in a dataset of 912 characters of which 53 (5·8%) were parsimony-informative. Finally, a concatenated set of sequence data was prepared and resulted in a dataset of 2252 characters of which 121 (5·4%) were parsimony-informative. Parsimony analysis resulted in two best trees with 432 steps (CI =  0·71, RI = 0·90) (Fig. 2). Both distance and parsimony analyses on a single sequence or concatenated set of sequences resulted in trees of similar topology, and similar to those published by Vincenot et al. (2008), with high bootstrap support for all nodes. It further substantiated the close phylogenetic relationship observed between L. biglobosa‘canadensis’ and L. biglobosa‘occiaustralensis’ (Vincenot et al., 2008). Following these analyses, all 23 L. biglobosa isolates could clearly be assigned to a previously described L. biglobosa subclade. All the isolates from central Canada, whatever their origin, could be assigned to L. biglobosa‘canadensis’ and showed very limited nucleotide polymorphism compared to a previously published reference sequence of Canadian isolates (Fig. 2). By contrast, the two isolates from Chile were L. biglobosa‘occiaustralensis’ isolates (Fig. 2). In Georgia, three subclades of L. biglobosa were represented: three L. biglobosa‘brassicae’ isolates were found in the Plains site, with very limited nucleotide polymorphism compared to published sequences from European or Chinese isolates; and three L. bigolobosa‘occiaustralensis’ isolates and one L. biglobosa‘canadensis’ were obtained from Griffin (Fig. 2).

Figure 2.

 One of two most parsimonious trees generated with concatenated ITS1-5·8S-ITS2, actin and β-tubulin sequence data. Leptosphaeria maculans‘brassicae’ is included as an outgroup. Bootstrap values (1000 replications) are indicated as percentages above the nodes. The length of the tree is 432 steps (CI = 0·71, RI = 0·90). Sequence data for the reference isolates (LbThlIBCN65, LbBraIBCN89, LbBraIBCN93, LbCanIBCN63, LbCanUnity, LbCanIBCN82, LbAusIBCN29, LbAusIBCN91, LbEryIBCN83 and LmBraLeroy) are from Mendes-Pereira et al. (2003), Vincenot et al. (2008) and Voigt et al. (2005). Ge: Georgia, Ch: Chile, Al: Alberta, Sk: Saskatchewan, Mb: Manitoba, A: Western Australia.

Race definition in L. maculans

The use of the typical differential set comprising Westar, Columbus, Bristol, Jet Neuf, 03-22-3-1, 01-23-2-1, Darmor-MX and Goéland was adequate to discriminate race structure for all populations showing similarity to European races, i.e. when isolates were depleted in avirulence alleles AvrLm2, AvrLm3 and/or AvrLm9. In particular, all isolates could be easily characterized at the AvrLm3, AvrLm7 and AvrLm9 loci because plant genotypes were available with only the corresponding Rlm gene (Table 2).

One ambiguity regarded characterization of AvrLm4 when using Jet Neuf (Rlm4) as a differential. The AvrLm4-Rlm4 interaction phenotype is that of a very typical gene-for-gene interaction, with only expression of a very small necrotic reaction at the inoculation point (Balesdent et al., 2001). By contrast, in a series of characterizations with Jet Neuf, resistance was expressed as a delayed, large-sized necrotic reaction (data not shown). To define more precisely the differences between these two interaction phenotypes, another Rlm4 genotype, Pixel, was used but, in a few cases, ambiguities remained. Finally, PCR amplification of the AvrLm4-7 gene generally confirmed that atypical resistance reactions are independent of the presence of the AvrLm4 allele. The final decision was to assign the AvrLm4 genotype only to those isolates expressing the typical AvrLm4-Rlm4 interaction phenotype and amplifying the AvrLm4 allele, the ‘atypical’ phenotypes in the absence of gene amplification could be due to a novel gene-for-gene interaction or to ill-defined non-specific resistance.

To discriminate between Avr alleles when the differential plant genotype harboured more than one corresponding R gene, different strategies were used. Usually isolates were avirulent (at least) against cvs Bristol, Darmor-MX and Goéland. In this case, it was difficult to identify the presence of AvrLm2 and AvrLm6 due to the presence of Rlm9 in both Bristol and Darmor-MX (Table 2). Since very few lines currently harbour Rlm6, and only in combination with other Rlm genes, the only way to discriminate the presence of AvrLm6 in isolates also harbouring AvrLm9 was by PCR-based assays showing the presence or absence of the avirulent allele with a 92·75% confidence (Fudal et al., 2009). To identify the presence of the currently uncloned AvrLm2, the first step was to use a new differential, Grizzly, harbouring both Rlm2 and Rlm3. However, due to the presence of AvrLm3 in some of the isolates, the last step to decide the presence of AvrLm2 was to combine the analysis with inoculation onto Heaven harbouring Rlm2 and Rlm7. Only 27 isolates harbouring AvrLm3, AvrLm7 and AvrLm9 could not be characterized at the AvrLm2 locus using this strategy.

Finally, one important ambiguity regarded characterization of AvrLm1. The resistance response of Columbus could not allow distinction between the presence of AvrLm3 alone or in combination with AvrLm1 when these isolates induced resistance on 03-22-3-1. This was partly resolved by using Cooper (Rlm1, Rlm4) and 04-35-2-5 (or 00-136-2-1) (Rlm1, Rlm7). The use of PCR markers showing a 98·3% efficiency for the diagnosis of the presence of a functional allele at the AvrLm1 locus (Gout et al., 2007) showed an unexpected level of complexity: whereas the amplification data were consistent with the interaction phenotypes observed on 04-35-2-5 and 00-136-2-1, numerous isolates were avirulent toward Columbus and/or Cooper (in the absence of AvrLm3 and/or AvrLm4) but had no AvrLm1 allele detectable by PCR, meaning complete absence of the gene (Gout et al., 2007). Following PCR amplification of the AvrLm1 locus in all of the isolates collected, the decision was made that all cases where resistance interaction phenotypes were observed towards Columbus/Cooper in the absence of amplification of AvrLm1 would be considered attributable to a novel putative avirulence gene termed here AvrLmx. The genetic characterization of this novel interaction is still to be done. This gene could be characterized by cotyledon-inoculation testing in only part of the population, i.e. in isolates lacking both AvrLm1 and AvrLm3, due to the presence of the corresponding Rlm genes in Columbus. Some populations in which AvrLm1 was very common such as populations from Western Australia, Mexico and Georgia thus could not be characterized at this locus (see below).

Avirulence alleles in American and Western Australian populations

In contrast to what has previously been observed for European isolates (Balesdent et al., 2006; Stachowiak et al., 2006), large variations between countries were observed for all investigated avirulence alleles. Moreover, most of the avirulence alleles showed a very large site-to-site variation within the central Canadian provinces.

AvrLm1 was absent or very rare in some locations (Chile, Ontario, Alberta and Saskatchewan), whereas it was still an important part of the populations in Georgia, Western Australia and Mexico (Fig. 3). AvrLm2 was absent or rare in Chile, Georgia, Western Australia and Ontario (Fig. 3; Table 3). It was prevalent in Mexico, as well as in Saskatchewan and Manitoba with site-to-site variations (Table 3). AvrLm3 showed a distribution similar to that of AvrLm2 except that it was absent from Mexican isolates (Fig. 3) and common in the limited set of isolates from Alberta analyzed here (Table 3). AvrLm4 was extremely variable between populations and showed a contrasted occurrence between two sampling locations within a single state/province (Table 3). AvrLm6 and AvrLm7 were prevalent in all populations from Mexico, Georgia, Chile, Ontario and Western Australia (82–100% of the isolates) (Fig. 3). AvrLm6 was also globally prevalent in central Canada, whereas AvrLm7 was very variable in central Canada (Table 3). AvrLm9 was rare or absent from Chile, Georgia, Ontario, Mexico and Western Australia (Fig. 3), whereas it showed large site-to-site variation in central Canada. Finally, the novel putative AvrLmx was absent in Chile and rare in Ontario whereas it was variable in central Canada (Table 3). It was also absent in the few isolates from Georgia and Western Australia lacking both AvrLm1 and AvrLm3 (data not shown).

Figure 3.

 Avirulence allele (AvrLm) occurrence in Leptosphaeria maculans populations. Data are percentages of isolates avirulent for each of the analyzed loci as identified following cotyledon-inoculation testing on a series of differential cvs and PCR amplification of AvrLm1, AvrLm4 and AvrLm6. AvrLmx was only characterized for isolates lacking AvrLm1 and AvrLm3.

Table 3.   Investigation of the frequency of each avirulence allele from different locations. Data are expressed as a ratio of isolates of Leptosphaeria maculans in the population harbouring a given avirulence locus.
No. of
AvrLm1 AvrLm2 AvrLm3 AvrLm4 AvrLm6 AvrLm7 AvrLm9 AvrLmx
  1. aIsolates obtained from the resistant Surpass 400 cv. in Chile are not included in the Table.

  2. bnd, not determined.

Elm CreekManitoba2501000769696448
Birch HillsSaskatchewan392·610028·246·289·756·441nd
Wongan hills,
Eradu, Esperance,
Merredin, Katanning
Avondale, Newdegate

Regional differences for individual alleles in Saskatchewan and Manitoba, Canada

Site-to-site variation in individual avirulence alleles was investigated in Manitoba and Saskatchewan, but not in Alberta due to the limited number of isolates that could be sampled in this province (Table 1). With the main exception of AvrLm6, AvrLm1 and AvrLm2 for Saskatchewan only, all other avirulence alleles showed a very large site-to-site variation, even when these were only separated by 20 to 50 km (Table 3). The five locations in Southern Manitoba, Letellier, Morris, Roland, Carman and Somerset, were exemplary of this variability with, for example, the ratio of AvrLm1 isolates ranging from 0% in Morris to 62·5% in Roland, 40 km away. Similarly 100% of the isolates harboured AvrLm2 in Morris versus only 48% in Roland (Table 3). In Morris, 79·5% of the isolates carried AvrLm3 versus only 20% in Letellier, 35 km away, whereas 18% of the Morris isolates carried AvrLm4 versus 70% in Letellier. Some locations like Roland showed specific features such as a much higher percentage of AvrLm1 and a much lower percentage of AvrLm2 isolates than all other locations in Manitoba or Saskatchewan (Table 3). Elm Creek, Manitoba, also had a higher ratio of AvrLm7 and lower ratio of both AvrLm9 and AvrLm3 isolates than all other locations in central Canada (Table 3). This resulted in large differences in the occurrence of Avr alleles throughout Saskatchewan and Manitoba: 0–62·5% for AvrLm1, 48–100% for AvrLm2, 0–84% for AvrLm3, 0–76% for AvrLm4, 0–80% for AvrLm7, 4–100% for AvrLm9 and 0–56% for AvrLmx (Table 3).

Race structure

Seven avirulence alleles were polymorphic in Western Australia. A total of 128 races (i.e. 27 combinations) could theoretically exist. Of these, only 10 races were observed in the sampling of 89 isolates, of which one, race Av1-(5)-6-7-(8) represented the vast majority (74.15%) (data not shown). The second most frequent race, Av1-(5)-6-(8) only represented 9% of the sampling, whereas all other races were only represented by 1 to 3 isolates.

In the case of American populations, where eight polymorphic AvrLm loci could be analysed in some locations, 256 races resulting from the combination of the alleles could theoretically exist. Forty-three races were identified among the 909 isolates for which AvrLm1–4, AvrLm6, AvrLm7 and AvrLm9 could be unequivocally identified. Of these, 34 could not be resolved at the AvrLmx locus. Thirty-two races were rare and only corresponded to a very limited number of isolates (nine or less) (Table 4). These were present at 11 distinct locations, most of them being in central Canada, with the main exception of race Av1-(5)-7-(8)-(x) represented by one isolate obtained in Chile and race Av3-4-(5)-6-(8)-(x) represented by one isolate from Ontario (Table 4). Compared to Australia, no race was overwhelmingly dominant, with the most common race, Av2-3-(5)-6-(8)-9-(x), present in 27·7% of the isolates (most of the isolates from central Canada, along with seven isolates from Ontario) and a total of nine races amounting to 80% of the isolates.

Table 4. Races of Leptosphaeria maculans identified in the American collection of isolates
RacePGaNo. of isolatesbOriginc
  1. aPG, Pathogenicity Grouping as discriminated by inoculation on a set of three cvs, Westar (susceptible control), Glacier and Quinta. The PG2 (avirulent on Quinta and Glacier), PG3 (avirulent on Quinta only) and PG4 (virulent on all three cvs) classification is according to Mengistu et al. (1991). The PGT (intermediate/avirulent on Glacier only) classification is according to Chen & Fernando (2006). The PG transcription from the race data is done on the assumption that the Glacier and Quinta genotypes used by Mengistu et al. (1991) and Chen & Fernando (2006) harboured Rlm2 and Rlm3 (Glacier), and Rlm1 and Rlm4 (Quinta) (Balesdent et al., 2001, 2002), and on the second assumption that none of these plant genotypes recognize AvrLmx (currently unknown).

  2. bOnly isolates that could be unequivocally characterized at all Avr loci (except for AvrLmx) are included in the Table.

  3. cGeorgia: (Gr: Griffin, Pl: Plains), Chile: (Te: Temuco), Ontario: (Ki: Kincardine), Manitoba: (Mo: Morris; So: Somerset; El: Elm Creek; Br: Brandon; Le: Letellier; Rol: Roland; Car: Carman), Saskatchewan: (Ros: Rosthern; La: Lanigan; Ba: Battleford; Ab: Aberdeen; Hu: Humbolt; Bi: Birch Hills; Me: Melfort; Sa: Saskatoon), Alberta: (Cam: Camrose; We: Wetaskiwin; Ft: Fort-Saskatchewan), Mx: Mexico.

Av1-2-3-(5)-6-(8)-9-(x)PG23Car, Sa, Br
Av1-2-4-(5)-6-7-(8)-(x)PG25Car, Rol, So
Av1-2-4-(5)-6-7-(8)-9-(x)PG22Rol, Bi
Av1-3-(5)-6-(8)-9-(x)PG27Rol, Car
Av2-3-4-(5)-6-(8)-9-(x)PG22Ros, Bi
Av2-3-4-(5)-6-7-(8)-(x)PG22Car, Bi
Av2-4-(5)-6-7-(8)PG295Le, Ab, Hu, Mo, El, Ros, So, Car, Rol, Sa
Av2-4-(5)-6-7-(8)-9PG27Sa, Car, Ros, Le
Av2-4-(5)-6-7-(8)-9-xPG23Rol, Sa, El
Av2-4-(5)-6-7-(8)-xPG254Ft, Me, Ba, Car, Bi, El
Av1-(5)-6-7-(8)-(x)PG361Gr, Pl, Te, Rol
Av1-4-(5)-6-7-(8)-(x)PG354Gr, Pl, Car, Rol, Sa, Le, So
Av4-(5)-6-7-(8)PG361Gr, Pl, Te, Ki
Av4-(5)-6-7-(8)-xPG32Car, Rol
Av2-(5)-(8)-9-xPGT2Bi, Br
Av2-(5)-6-(8)-9PGT25La, Le, Ros, Bi
Av2-(5)-6-(8)-9-xPGT5Br, Bi
Av2-(5)-6-7-(8)PGT34Bi, So, Hu, Ab, El
Av2-3-(5)-(8)-9-(x)PGT9So, Br, Ros, Sa, Bi, Hu
Av2-3-(5)-6-(8)-(x)PGT6Car, Rol
Av2-3-(5)-6-(8)-9-(x)PGT252Le, Br, Car, Sa, We, Ft, Me, La, Ba, Ab, Hu, Mo, Rol, Ros, So, Cam, Ki
Av(5)-6-7-(8)PG4158Gr, Pl, Te, Ki

No isolate was fully depleted in avirulence alleles, with two resolvable avirulence alleles the minimal number observed, and no isolate showed the whole complement of possible avirulence alleles, with six being the maximal number observed.

One main difference throughout the American continent was the number of virulences per isolate. For this criterion, a significant separation was obtained between isolates from Georgia, Chile and Ontario on the one side, and those from central Canada on the other (F = 14·25, P < 0·0001) (Table 1). In addition, not only the number of virulence(s), but also the combination of virulence alleles showed features often specific to one country/province. As a consequence, only one race, Av1-(5)-6-7-(8)-(x), representing 6·7% of the isolates, is found throughout the American continent (Chile, Georgia and one location in central Canada) (Table 4). A second one is found in Georgia and diverse locations in central Canada, and two more were found in Ontario, and these plus a third making up 30·8% of the isolates were found in Chile and Georgia (Table 4). Only one race is found throughout Canada (Table 4). By contrast, only one race is found in Mexico (and in no other American location) and 35 races are specific to central Canada and not found elsewhere in the continent (Table 4). Finally, due to the large site-to-site variation observed in central Canada, and even though many more races were found in Manitoba than in other locations, no significant differences between countries/provinces were observed for the number of races (F = 1·21, P = 0·35), or the richness in races as estimated by the Margalef Index (F = 2·36, = 0·10) or the Simpson diversity index (F = 0·90, = 0·49) (Table 1).

Distance analyses based on the seven avirulence alleles AvrLm1, AvrLm2, AvrLm3, AvrLm4, AvrLm6, AvrLm7 and AvrLm9, and including previously published data on European isolates, clearly discriminated three clusters of isolates: the first one included all European isolates, along with Chilean isolates and isolates from Ontario. It had a very short branch distance indicative of a low diversity (Fig. 4). The second one grouped all Western Australian isolates with those from Georgia. Longer branch distances compared to the first cluster are indicative of higher race diversity. The third cluster, clearly separated from the two others, grouped all isolates from central Canada with long branch distances indicative of extreme race diversity from location to location.

Figure 4.

 Neighbour-joining analysis of populations of Leptosphaeria maculans based on avirulence (Avr) allele data (AvrLm1-AvrLm4, AvrLm6, AvrLm7 and AvrLm9). Isolate numbers have suffix letters referring to their geographic origin (Au: Australia, Ch: Chile, Fr: France, Ge: Georgia, Gr: Germany, Mb: Manitoba, On: Ontario, Po: Poland, Sk: Saskatchewan, Sw: Sweden, UK: United Kingdom) followed by the name of the location of isolation. Data for European isolates are from Balesdent et al. (2006) and Stachowiak et al. (2006).

Race structure at the regional scale

An extreme race diversity from location to location was observed at the country/province scale in central Canada, resulting in non-significant differences between regions with regards to richness in races (Table 1). Some locations such as Birch Hills in Saskatchewan, and Carman and Roland in Manitoba, showed a large diversity of races, with up to 14 races present at Roland (Table 1). This was variable from one location to another in the same province, with 3 to 14 races per location in Manitoba and 2 to 10 races per location in Saskatchewan (Table 1). As also observed to a lesser extent for populations from the two Georgia locations, race structure at specific sites in central Canada was unrelated to geographic proximity between sites with e.g. populations sampled in sites located a few km apart being less related to one another than to populations from another province (Fig. 4). In addition, the wide site-to-site variation resulted in the presence of races unique to one given location. This was mostly the case for fields in Manitoba, with all of the seven investigated locations showing the presence of at least one race unique to each location (Table 4). By contrast rare races, represented by only a few isolates, were often found in distantly separated locations, e.g. race 1-2-3-(5)-6-(8)-9-(x) represented by three isolates from Saskatoon (Saskatchewan), Carman and Brandon (Manitoba) (Table 4).


This first wide-scale survey of the L. maculans-L. biglobosa species complex in the American continent showed a series of specific features. Among these, the following are of particular relevance: (i) the firm establishment of L. biglobosa‘canadensis’ as a major sub-clade widely distributed throughout Canada; (ii) the diversity of L. biglobosa sub-clades found at the continent scale and the presence of up to three different sub-clades at the scale of a single USA state; (iii) the large diversity of L. maculans races throughout the American continent; (iv) the close relationship observed between race structure of European isolates with that of Eastern Canada, Chile, Georgia and Western Australia; and (v) in contrast, the race structure of central Canadian L. maculans isolates.

In all countries, except China where only L. biglobosa has been found to date, L. maculans and L. biglobosa are closely associated in oilseed rape fields, being found together in the same field and often within tissues of the same individual plants (Jedryczka et al., 1999; West et al., 2001; Vincenot et al., 2008). The results presented here show the first identification of L. biglobosa in Chile and in the South-Eastern USA and indicate the constant association of L. biglobosa with L. maculans in oilseed rape fields, independent of the L. biglobosa subclades. In addition to the previously described association between L. maculans and L. biglobosa‘brassicae’ in the EU (Mendes-Pereira et al., 2003), L. maculans and L. biglobosa‘occiaustralensis’ in Western Australia (Vincenot et al., 2008) and L. maculans and L. biglobosa‘canadensis’ in New South Wales (Australia) (Van de Wouw et al., 2008), the data confirm such an association of L. maculans with L. biglobosa‘canadensis’ throughout Canada and in Georgia, with L. biglobosa‘brassicae’ in Georgia, and with L. biglobosa‘occiaustralensis’ in Georgia and Chile.

Six subclades of L. biglobosa are reported to date (Vincenot et al., 2008), but it is questionable whether these are representative of the natural occurrence or distribution of L. biglobosa since only two of them, L. biglobosa‘brassicae’ and L. biglobosa‘occiaustralensis’, were represented by an important number of isolates in international collections. It was dubious whether subclades obtained from cruciferous weeds, B. napus or B. juncea in Saskatchewan could be considered major clades of L. biglobosa, and Vincenot et al. (2008) questioned whether L. biglobosa‘canadensis’ could just be a rare variant of L. biglobosa‘occiaustralensis’. Along with the recent finding in New South Wales (Australia) of the occurrence of L. biglobosa‘canadensis’ (Van de Wouw et al., 2009), the data regarding the unique and general presence of L. biglobosa‘canadensis’ throughout central Canada firmly establish this as one of the three major L. biglobosa subclades. In addition, L. biglobosa‘canadensis’ seems to still represent an important part of the population in Alberta. Leptosphaeria maculans was reported for the first time in 1983 in Alberta and does not seem to be firmly established in this province, thus explaining the difficulty observed here and by others (J. Davey, unpublished data) to isolate enough L. maculans isolates from this province. Similarly, the finding of L. biglobosa‘occiaustralensis’ outside of Western Australia establishes this subclade as a widely distributed one, whose diversity and migration patterns are now worth investigation.

One surprising finding of this work was the wide diversity of L. biglobosa in America. To date, L. biglobosa‘brassicae’ was regarded as a highly monomorphic sub-clade present throughout the indo-european continent (from Western Europe to China), and no other sub-clade of L. biglobosa was described at this scale. Similarly, in accordance with data on collection of isolates (Mendes-Pereira et al., 2003) the data indicate the uniqueness of the highly monomorphic L. biglobosa‘canadensis’ sub-clade on oilseed rape throughout central Canada. By contrast, three L. biglobosa sub-clades are found in Australia, of which one, L. biglobosa‘australensis’ seems to be very rare (Van de Wouw et al., 2008; Vincenot et al., 2008). Similarly, the three main L. biglobosa sub-clades are found at the American continent scale, two of them, L. biglobosa‘occiaustralensis’ and L. biglobosa‘canadensis’ being found in a single field in Georgia. The reasons for this high level of diversity in Georgia are unknown but, since seeds readily transport L. biglobosa (Fitt et al., 2006), it can be hypothesized that it is linked with the recent establishment (at the end of 1980s) of oilseed rape/canola as a crop in Georgia, using plant genotypes originating from Europe, Australia and Canada, and joint breeding programmes developed between Western Australia and Georgia (Phillips et al., 1999; D. Phillips and M. Barbetti, unpublished data). A non-exclusive alternative explanation would be that the recently established oilseed rape cropping still faces diverse indigenous fungal species present on diverse crucifers, including those widely grown in the region in family gardens (D. Phillips, unpublished data), that may have been eradicated in intensive and older oilseed rape cropping systems.

Leptosphaeria maculans large-scale race structure analyses were previously performed at the European Union (EU) scale using both an isolate trapping protocol and the systematic phenotyping of isolates at each resolvable Avr loci (Balesdent et al., 2005, 2006; Stachowiak et al., 2006). Here, the same trapping protocol as the one used for analysis of EU isolates was used, and uneasy-to-characterize Avr loci were resolved by combining inoculation tests and PCR-based approaches. This is thus the first report of race structure analysis in Chile, and also the first systematic analysis in Australia, Mexico, Georgia, Ontario and central Canada.

In the EU, a very low level of race diversity was observed at the continent scale with complete lack of avirulence alleles AvrLm2, Avrlm9, and to a lesser extent, AvrLm3, and prevalence of the two avirulence alleles AvrLm6 and AvrLm7 (Balesdent et al., 2006; Stachowiak et al., 2006). Retrospective analysis of data showing the lack of a resistance reaction on Columbus in the absence of AvrLm1 and AvrLm3 (Balesdent et al., 2006; Stachowiak et al., 2006) also suggests EU populations do not harbour AvrLmx. Populations from Chile and Georgia showed features very similar to those of EU populations. These have a complete lack of avirulence alleles AvrLm2, AvrLm9, AvrLm3, and when possible to evaluate, AvrLmx, and the prevalence of the two avirulence alleles AvrLm6 and AvrLm7. These features were also found to a lesser extent within populations from Western Australia and Ontario. In Chile and Georgia, this resulted in a very limited number of combinations of virulences and low race richness. Similarly in Ontario, 80% of the isolates belonged to a single race, although all Avr loci were polymorphic except AvrLm1 and AvrLm6. In Western Australia, although one race was prevalent among the 10 identified, richness of races was higher in Western Australia than in the EU, Georgia, Ontario or Chile, with all Avr loci being polymorphic for these populations. In this respect, races harbouring the highest number of virulences of the sampling (i.e. race Av3-(5)-(8)-(x) and race Av(5)-7-(8)) were found in Western Australia. These data are consistent with previous analyses performed on a very limited set of collection isolates, obtained from uncharacterized plant genotypes, and showing that four Ontario isolates were closely related to European races, whereas a wide diversity of races, including the richest available race, were found in a collection of 23 Australian isolates of which 18 were from Western Australia (Balesdent et al., 2005). These data are less easy to relate to the limited data available for Georgia (Phillips et al., 1999; Pongam et al., 1999) since isolates were characterized using PG (Pathogenicity Group) terminology, using the Westar-Quinta-Glacier differential set which does not identify Avr alleles such as AvrLm6, AvrLm7 or AvrLm9 (Balesdent et al., 2005).

By contrast to these populations showing relatedness with EU populations, two very specific populations were characterized here: those from Mexico and those from central Canada. The isolates obtained from cabbage in Mexico were remarkable in their lack of diversity (only one race represented) and their pattern of avirulence alleles. The race they belong to is not found in other parts of the American continent nor in the rest of the world. Mexican isolates from cabbage were previously characterized as ‘PG2’ (Moreno-Rico et al., 2001), which is consistent with the 100% presence of AvrLm1 and AvrLm2 within these isolates. However, these data on Mexican populations have to be taken with care since only 28 isolates were analyzed here, from only two locations. More extensive samplings and analyses are currently in progress to assess whether the sample is representative of current Mexican populations that could be extremely isolated from those of the rest of the American continent.

The previously published works on central Canadian provinces (and North Dakota) indicated the prevalence of PG2 isolates (harbouring AvrLm2 and/or AvrLm3 and AvrLm1 and/or AvrLm4) throughout central Canada with some recent surges of PGT, PG3 and PG4 isolates (e.g., Chen & Fernando, 2006) (Table 4). If the Quinta genotype used by Chen & Fernando (2006) harbours Rlm4, as is often the case (Balesdent et al., 2001), and suggested by the similar reaction of Quinta and Dunkeld (known to harbour only Rlm4; Rouxel et al., 2003b) to PG3 isolates (Dusabenyagasani & Fernando, 2008), the data here confirms the importance of PG2 isolates throughout Canada with 33% attributable to this pathogenicity group. However, the prevalent PG is PGT, represented by 59·2% of the isolates, whereas only eight PG3 are found in Manitoba, and none in Saskatchewan. PG4 are not found here outside of Ontario, in contrast to that described by Chen & Fernando (2006). However, as already observed by Balesdent et al. (2005), it mainly stresses the inadequacy of the PG terminology, with for example, 21 different races that can be related to PG2. More importantly, use of the PG terminology has the main disadvantage of hiding the wide diversity of races present in central Canada, and the strong site-to-site variability in both race richness and content.

A strong diversity of populations is usually associated with their belonging to a centre of origin of the species, where they co-exist and co-evolve along with wild hosts (Stukenbrock & McDonald, 2008). However, the history of the spread of L. maculans contradicts this assumption. Leptosphaeria maculans is known to have been present in Europe for at least 80 years, with one isolate obtained in 1923 from cabbage being L. maculans (Mendes-Pereira et al., 2003), and the oldest establishment of a pathogen on stems of cabbages (termed at the time Sphaeria lingam) dating back from 1791 (Henderson, 1918). Similarly, there are indications that L. maculans was established for a long time in some parts of the USA such as Wisconsin, with damaging epidemics on cabbage fields observed at the beginning of the 1900s (Henderson, 1918). The presence of L. maculans in Australia is also postulated to be ancient, with a first report on cabbage in 1925 (Sivasithamparam et al., 2005). By contrast, the colonization of central Canada by L. maculans is thought to be very recent, and only L. biglobosa was present until the beginning of the 1970s (Gugel & Petrie, 1992). From an initial establishment in Saskatchewan during the mid-1970s (first record in 1975 in three widely separated fields in central Saskatchewan), probably from a single or very limited introduction (Gugel & Petrie, 1992; Kutcher et al., 1993), L. maculans spread throughout central Canada (Manitoba between 1978 and 1981, and then Alberta, close to the border with Saskatchewan, in 1983) and became predominant in the population (Gugel & Petrie, 1992) as is also shown in the present study. These data are contradictory with the hypothesis that central Canada could be the geographic origin of the fungus. In contrast, the countries where the fungus has been present for the longer time are also those with the most reduced race richness. In the case of European populations, this was explained by the continuous selection pressure exerted by major genes used in oilseed rape which tended to eradicate in turn avirulence genes AvrLm2, AvrLm3, AvrLm9 and explain the low amount of isolates still harbouring AvrLm1 and AvrLm4 (Balesdent et al., 2006). The similar race structure observed in Chile, Georgia and Ontario would then be due to recent introduction via global seed trade of EU/Australian isolates, due to the use of EU or Australian plant cultivars as crop or their inclusion in breeding schemes, and/or to exposure of the natural population to similar selection pressure as in the EU. Western Australian isolates would then have a similar history with a wider diversification due to the adaptation of the pathogen to cruciferous weeds harbouring resistance genes uncommon in oilseed rape (Sivasithamparam et al., 2005). The high level of diversity of central Canadian isolates would then be linked to the introduction of a core group of isolates which have not been submitted to the same Rlm selection pressure as EU isolates, and thus still harbouring AvrLm2, AvrLm3 and AvrLm9, but being mostly depleted in AvrLm7. The high level of diversification would thus be linked with natural variation in the absence of selection pressure since one of the most popular cultivars grown in Canada for years, Westar, is known to harbour no resistance gene. In the near future, this hypothesis will be evaluated using neutral markers and the development of migration analyses.

The data presented here have important consequences in terms of breeding for resistance and durable management of resistance. Strategies and world-wide breeding can be transposed from the EU to Chile, Ontario, Georgia or, to a lesser extent Australia, with some major genes still showing an adequate level of resistance against most of the populations. The situation is much more complex for central Canada, and suggests that specific breeding and management of resistance strategies need to be adopted in this region. It must be stressed that the PG2 terminology often disguised Canadian populations as only consisting of weakly pathogenic isolates. However, the common occurrence of AvrLm2, and much rarer occurrence of AvrLm7, suggest that old Rlm genes can indeed be recycled from EU-originating oilseed rape varieties, but the currently successful strategy of breeding with Rlm7 resistance in Europe is pointless for Canada. Most interestingly, the main concern here is the diversity of races from one location to the other in a single province, even a few kilometres apart. Whereas strategies for deployment of resistance are straightforward throughout the EU with a high homogeneity of race structure even from one country to the other, it will be extremely challenging to successfully deploy major gene resistance within central Canada.


The authors wish to thank G. Leroy, L. Coudard (INRA Bioger, Versailles, France) for plant management; J. P. Narcy and J. Roux (INRA Bioger, Versailles, France) for technical assistance and A. Gautier (INRA Bioger, Versailles, France) for sequencing. We also thank R. Delourme (INRA Rennes, APBV) for providing seeds of Darmor MX.