Disease development and genotypic diversity of Puccinia graminis f. sp. avenae in Swedish oat fields


  • A. Berlin,

    Corresponding author
    1. Department of Forest Mycology and Plant Pathology, Swedish University of Agricultural Sciences, Box 7026, 750 07 Uppsala
    Search for more papers by this author
  • B. Samils,

    1. Department of Forest Mycology and Plant Pathology, Swedish University of Agricultural Sciences, Box 7026, 750 07 Uppsala
    Search for more papers by this author
  • A. Djurle,

    1. Department of Forest Mycology and Plant Pathology, Swedish University of Agricultural Sciences, Box 7026, 750 07 Uppsala
    Search for more papers by this author
  • H. Wirsén,

    1. Department of Forest Mycology and Plant Pathology, Swedish University of Agricultural Sciences, Box 7026, 750 07 Uppsala
    2. Swedish Rural Economy and Agricultural Societies, Halland, Lilla Böslid 146, 305 96 Eldsberga, Sweden
    Search for more papers by this author
  • L. Szabo,

    1. US Department of Agriculture–Agricultural Research Service, Cereal Disease Laboratory, University of Minnesota, St Paul, MN 55108, USA
    Search for more papers by this author
  • J. Yuen

    1. Department of Forest Mycology and Plant Pathology, Swedish University of Agricultural Sciences, Box 7026, 750 07 Uppsala
    Search for more papers by this author

E-mail: anna.berlin@slu.se


The disease development and population structure of Puccinia graminis f. sp. avenae, which causes stem rust on oats, were studied to investigate if sexual reproduction plays an important role in the epidemiology of the disease. The genetic population structure of P. graminis f. sp. avenae in Sweden was investigated by sampling 10 oat fields in July and August 2008 and seven fields during the same period in 2009. Nine single-pustule isolates were first used to test simple sequence repeat (SSR) markers developed for P. graminis f. sp. tritici. Eleven of the 68 tested SSR markers were useful for genotyping P. graminis f. sp. avenae. For the main study, DNA from single uredinia was extracted and the SSR markers were used to genotype 472 samples. Both allelic and genotypic diversity were high in all fields, indicating that P. graminis f. sp. avenae undergoes regular sexual reproduction in Sweden. No significant relationship between genetic and geographic distances was found. Disease development was studied on two farms during 2008 and 2009. The apparent infection rates ranged between 0·17 and 0·55, indicating the potential for rapid disease development within fields. The incidence of oat stem rust has increased recently in Sweden. One possible explanation is a resurgence of its alternate host, barberry (Berberis spp.), after the repeal of the barberry eradication law in 1994. Barberry is present in several grain-producing areas in Sweden, which supports the conclusion that P. graminis f. sp. avenae undergoes regular sexual reproduction there.


Stem rust, caused by Puccinia graminis, is a serious disease of cereal crops. Puccinia graminis is an obligate biotrophic fungus and its urediniospores have been reported to infect 365 different grass species (Leonard & Szabo, 2005). Puccinia graminis is a heteroecious rust with five spore stages, and the alternate host Berberis spp. (barberry) is needed for the pathogen to complete sexual reproduction. This species is divided into different formae speciales (f. sp.), each being adapted to a specific host or group of hosts. Oat (Avena sativa) is infected by P. graminis f. sp. avenae. Puccinia graminis, like many other fungal pathogens that have a sexual stage, is asexual for most of its life cycle.

The disease progress of stem rust is often rapid and it may result in severe yield losses if the crop is infected at booting or heading (DC 45–59, Zadoks et al., 1974). If the onset of infection is late, in the soft dough to mature stages (DC 71–83), the losses will be negligible (van der Plank, 1963). The time of infection is therefore important in order to be able to estimate the potential damage of the disease and to decide if fungicide treatment is economically justified. Urediniospores of Puccinia spp. can be transported long distances in the atmosphere, making the disease a threat for large cereal-growing areas (Kolmer, 2005). On a local scale, and if the alternate host is present, the primary infection will occur earlier than if the infection is initiated by spores arriving via long-distance dispersal. The genetic diversity is also greater in areas where the alternate host is in close proximity to the cereal or grass host. The relative importance of early and late infections has been investigated for powdery mildew (O’Hara & Brown, 1996), where it was shown that once the disease is established in a field, the proportion of immigrants among sampled spores drops and the proportion originating from within the field increases.

Oat is the third most common cereal crop in Sweden, but it is only spring-sown and it is primarily used as fodder (Fogelfors, 2001). The incidence of oat stem rust has increased during the last decade and yield losses up to 2 t ha−1 (or 30%) have been reported from untreated field plots compared with treated ones, although disease onset is often reported late in the growing season (Mellqvist & Waern, 2010). This has led to a practice where Swedish farmers use more fungicides to control the disease, partly for increasing grain yield, but also to improve the quality of the straw. The alternate host Berberis vulgaris (common barberry) is present in cereal-growing areas of Sweden. The Swedish legislation regarding barberry eradication was repealed in 1994, and since then a trend of increasing barberry occurrence (Georgson, 1997; Rydberg & Wanntorp, 2001; Bertilsson et al., 2002; Fröberg, 2006; Edqvist & Karlsson, 2007; Tyler, 2007) has been seen and aecia have been observed on barberry.

Southern Sweden has a warm, fully humid, temperate climate (Kottek et al., 2006), which means that the temperature fluctuates around freezing during the winter. No oats are grown between harvest in August/September and planting in April/May and thus there is no green bridge on which the urediniospores can overwinter. The production of oats south of the Baltic Sea is limited, which means that there are few hosts on which urediniospores of P. graminis f. sp. avenae may overwinter in that area and subsequently be dispersed to Sweden. This implies that infections start either from aecial infections on local barberry, or from urediniospores immigrating from other regions where the pathogen is able to overwinter on oats or wild oats (Avena spp.).

The presence of the alternate host enables P. graminis to complete its sexual cycle, which allows for recombination of factors leading to virulence and thus increases the evolutionary potential of the pathogen to overcome resistance in the host population (McDonald & Linde, 2002). Genetic variation within populations of pathogens may also be created by mutation, migration of individuals, as well as genetic drift and extinction events. Simple sequence repeat (SSR) markers are common in the study of the biology of plant pathogens and have previously been used to study the population structure of Puccinia graminis f. sp. tritici (Szabo, 2007; Jin et al., 2009; Zhong et al., 2009; Admassu et al., 2010). Apparently, only one study using sequence-tagged microsatellite markers on P. graminis f. sp. avenae has been reported, from Australia, which showed that the oat stem rust population is highly clonal on that continent (Keiper et al., 2006). The main objective of the present study was to investigate the population biology of the stem rust fungus on oats in Sweden in order to determine the importance of sexual reproduction for the spread and development of stem rust in commercial oat fields.

Materials and methods

All samples used were collected from plots in commercial oat fields, which were not treated for pests and diseases. Stem rust fungal samples (P. graminis f. sp. avenae) were collected from 10 fields in July and August 2008 and seven fields during the same period in 2009 (Fig. 1). In each field, infected stem samples were collected within a regular grid design of 30 × 100 m. The samples were taken along three rows, 10 m apart, and 10 samples were taken in each row at a distance of 10 m, resulting in 30 samples from each field. Each sample was kept in a separate paper bag. All oat fields were moderately to heavily infected with stem rust at the time of sampling. The fields were planted with five commonly used varieties. At all sampling sites, the close surroundings and the edges of the sampled fields were inspected for occurrence of Berberis spp.

Figure 1.

 Map of sample collection sites in Sweden. Numbers represent the oat fields in which the stem rust (Puccinia graminis f. sp. avenae) samples were collected. Letters represent the sites where disease development was studied: A, Fransåker; B, Stora Bärby. Bar = 100 km.

Disease development was studied during the growing seasons in three untreated survey field plots; at Fransåker in 2008 and 2009 and at Stora Bärby in 2009 (Fig. 1). Before disease onset, the fields were visited once a week. At each time point, 40–50 straws were randomly selected from different parts of the plot. The incidence and severity of disease was registered on three occasions in both seasons, where the incidence was recorded as the proportion of infected stems and the severity as the proportion of infected stem surface (James, 1974). The rate of disease progress, expressed as the apparent infection rate, r, was calculated by using the equation dxt/dt = rxt (1−xt), where xt is the disease proportion at time t. This equation assumes that disease increases according to logistic growth (Zadoks & Schein, 1979).

SSR marker selection

A subsample of nine single-pustule isolates of P. graminis f. sp. avenae collected from oat fields in 2008, and one reference isolate of P. graminis f. sp. tritici, were selected for testing simple sequence repeat (SSR) markers developed for P. graminis f. sp. tritici (Table S1). Procedures for extraction of DNA, PCR reactions and size determination of the products were performed as previously described by Szabo (2007). Most of the SSR markers tested have previously been used in studies of P. graminis f. sp. tritici (Szabo, 2007; Jin et al., 2009; Zhong et al., 2009), although other markers developed but not selected for final use in these studies were also tested (Table S2). SSR markers that were especially designed to include repeats of three base pairs were targeted for testing, as they are more robust than SSRs based on two base pair repeats. The SSR markers that produced amplicons were selected for future testing with an additional set of 118 single-pustule samples to confirm their usefulness.

DNA extraction and SSR analysis

For the main survey, single uredinia were cut out from collected stems and each transferred to a 2-mL plastic tube together with a 6-cm-long piece of air-dried oat leaflet. Thirty glass beads (2 mm in diameter) and approximately 0·4 mg of diatomaceous earth were put into each plastic tube. The samples were shaken twice for 30 s at a speed of 5000 rpm in a FastPrep shaker (Precellys24-Dual; Bertin Technologies). The DNA extractions were carried out using an OmniPrep kit (GenoTech) according to the manufacturer’s instructions for fungal tissue. The DNA concentration was measured using a spectrophotometer (ND-1000; NanoDrop) and the DNA samples were diluted to 20 ng μL−1 for further analyses.

Amplification of the SSR markers was carried out as a multiplex reaction in 10 μL 1× Phusion HF Buffer (Finnzymes), 0·08 mm dNTP, 0·2 μg μL−1 bovine serum albumin (Promega), forward primers at 0·5 μm each, either labelled with HEX or with FAM, the two corresponding reverse primers (Table 1) at 0·5 μm each, 20 ng DNA template and 0·2 U Phusion DNA Polymerase (Finnzymes). The thermocycling conditions were: denaturation at 98°C for 5 min, 35 cycles at 98°C for 30 s, 62°C for 30 s and 72°C for 30 s, with a final extension at 72°C for 10 min.

Table 1. Primer characteristics for the 11 simple sequence repeat (SSR) markers developed for Puccinia graminis f. sp. tritici, in this study used on Puccinia graminis f. sp. avenae
SSR locusPrimerPrimer sequence (5′–3′)ColourRepeat motif Na Size range (bp) Ho He Null allele frequencyaReference
  1. Na: number of observed alleles; Ho: observed heterozygosity (SE in parenthesis); He: expected heterozygosity (SE in parenthesis).

  2. aEstimate of null allele frequency by the EM algorithm of Dempster et al. (1977).

  3. bS. Zhong, North Dakota State University, Fargo, ND, USA.

Pgestssr02121AAGFGTTTGCCTGATGATGGATGAFAMAAG6232–2470·363 (0·031)0·515 (0·007)0·132Zhong, unpublished datab
PgtSSR2121FAAAATGATGGTCTCCTTGGCTAFAMTC-rich11159–1800·523 (0·039)0·524 (0·020)0·043 Szabo, 2007
Pgestssr02424FTCATCGACCAAGAGCATCAGHEXATG6121–1360·576 (0·042)0·599 (0·012)0·062Zhong, unpublished data
Pgestssr109109FCCATCCGATCATTTCTTCGTHEXAGG10152–1760·494 (0·035)0·477 (0·014)0·025 Zhong et al., 2009
Pgestssr171171FAGGCTCAACACCACCCATACHEXAGC28124–2430·447 (0·036)0·710 (0·022)0·183 Zhong et al., 2009
Pgestssr255225FCATCTGATCACCGTCACAGCFAMAAC14206–2510·651 (0·023)0·626 (0·024)0·054Zhong, unpublished data
Pgestssr279279FATCGAAGAGCCGTTCACTGTHEXATG7165–1830·452 (0·036)0·467 (0·023)0·028 Zhong et al., 2009
Pgestssr280280FGACTATCAACGGCTGGAAGGFAMAAC16141–1950·181 (0·023)0·682 (0·025)0·341Zhong, unpublished data
Pgestssr368368FCATCTGATCACCGTCACAGCFAMAAC10216–2430·598 (0·036)0·570 (0·010)0·031 Zhong et al., 2009
PgtCAA53CAA53FAGGCTCAACACCACCCATACFAMCAA9175–2140·048 (0·013)0·108 (0·030)0·101 Jin et al., 2009
PgtCAA93CAA93FCACTCTCGCCAAACCTCATTHEXCAA22144–2330·212 (0·044)0·513 (0·070)0·291 Jin et al., 2009

Amplification products (2·5 μL) were analysed on an agarose gel. The concentration of the amplification products was adjusted by adding up to 5 μL of distilled deionized water. Lengths of SSR amplicons were determined using an ABI 3730xl DNA Analyzer (Uppsala Genome Center at Rudbeck Laboratory, Uppsala University). The lengths of the amplicons were scored using the computer software GeneMarker (SoftGenetics) to identify the different genotypes. In the rare case where more than two alleles were present, the locus was coded as missing.

Data analysis

The usefulness of the primers (Table 1) was evaluated by calculating the observed (Ho) and expected (He) heterozygosity for each primer pair using GenAlEx v.6.4 (Peakall & Smouse, 2006). Ho was the proportion of samples that are heterozygous at the respective loci and He the proportion of samples within the loci expected to be heterozygous under random mating. The gene diversity was the number of alleles at each locus. The occurrence of null alleles was evaluated by using the EM algorithm (Dempster et al., 1977) in the software FreeNA (Chapuis & Estoup, 2007) with 1000 replicates for bootstrap calculations.

Single-uredinia samples from each collection site and year were grouped into one population, giving 17 populations in total (Table 2). Samples with more than four missing values were discarded. Genotypic diversity is reported as a G/N value, the number of genotypes divided by the total number of genotyped samples at a particular site. Observed (Ho) and expected (He) heterozygosity within populations were calculated using GenAlEx v.6.4 (Peakall & Smouse, 2006), where Ho was the average observed heterozygosity and He is the expected heterozygosity within a population with random mating. The inbreeding coefficient (FIS) was calculated using the equation proposed by Weir & Cockerham (1984) in the software Fstat v.2.9.3 (Goudet, 1995). Estimation of the selfing rate, s(g2), from the distribution of multilocus heterozygosity in each field population was performed in the software rmes, using 1000 iterations (David et al., 2007). This method has been shown to be insensitive to null alleles (David et al., 2007). To test for association between different loci, the index of association (IA) was calculated for each population separately, using the software MultiLocus v.1.3b (Agapow & Burt, 2001). Because of the high frequency of null alleles within particular loci, the IA tests for significance were carried out with the option ‘missing data fixed’ and replicated 1000 times for each population.

Table 2. Details of population genetic diversity in Puccinia graminis f. sp avenae at 11 microsatellite loci
PopulationLocationCultivarDay of collectiona N G/N Na FIS Ho He IA P
  1. N: number of samples; G/N: number of genotypes divided by number of samples; Na: number of observed alleles; FIS: inbreeding coefficient in relation to subpopulation; Ho: observed heterozygosity (SE in parenthesis); He: expected heterozygosity (SE in parenthesis); IA: index of association; and P: its P-value.

  2. aDay of collection, dd.mm.yy.

 1FransåkerBelinda01·08·08271·00620·4870·329 (0·062)0·618 (0·044)0·5090·563
 2IngvastaIvory06·08·08291·00450·3330·363 (0·074)0·526 (0·062)0·3410·045
 3EvertsholmIngeborg11·08·08300·93410·1330·472 (0·085)0·533 (0·057)0·1840·145
 4SkarpenbergaBelinda11·08·08281·00460·2500·388 (0·061)0·503 (0·045)0·1290·290
 5BettnaBelinda13·08·08281·00440·0570·470 (0·087)0·487 (0·065)0·1640·098
 6Bränne ÖvregårdSvava13·08·08300·90440·0540·436 (0·090)0·449 (0·074)−0·0030·485
 7PattalaBelinda13·08·08271·0038−0·0280·501 (0·091)0·477 (0·058)0·1080·125
 8UltunaIvory22·08·08221·00440·1670·413 (0·071)0·481 (0·063)0·0890·281
 9HjälmarsholmKerstin26·08·08261·00560·3330·350 (0·066)0·510 (0·062)0·0730·832
10GötalaIvory26·08·08251·00560·2440·443 (0·087)0·567 (0·054)−0·2151·000
11SkarpenbergaBelinda30·07·09281·00600·2770·419 (0·067)0·564 (0·061)0·3890·149
12StäholmKerstin04·08·09281·00500·2620·398 (0·059)0·524 (0·061)0·0240·649
13FransåkerBelinda04·08·09301·00610·3060·424 (0·050)0·593 (0·053)−0·0830·665
14IngvastaIvory05·08·09261·00550·4490·330 (0·047)0·578 (0·042)0·0630·661
15Bränne ÖvregårdKerstin06·08·09291·00550·2660·422 (0·047)0·560 (0·047)−0·0960·584
16Klostergården DalaBelinda07·08·09301·00380·1300·421 (0·067)0·474 (0·066)0·3320·002
17EvertsholmBelinda17·08·09291·00540·1640·438 (0·065)0·511 (0·067)−0·0490·852

To evaluate differences between the populations, varieties and years, analysis of molecular variance (amova) was performed using GenAlEx v.6.4. Using the same program, the correlated pairwise and overall theta (θ; estimation of Wright’s FST) and the corresponding migration rate (where Nm = [(1/θ) − 1]/4) were calculated based on the heterozygosity between individuals within each population (Weir & Cockerham, 1984) and the significance of each value was calculated by 9999 permutations. A Mantel test was performed to analyse the isolation by distance by testing the calculated pairwise genetic differences against the calculated pairwise geographical distances using GenAlEx v.6.4 (9999 permutations).


The first disease symptoms were detected as early as 10 July at Stora Bärby in 2009 and as late as 1 August at Fransåker in 2008 (Table 3). At Fransåker in 2008, precipitation was very low until the end of June, then there followed a dry period of 3 weeks. Rain occurred between the end of July and mid-August. At the same site in 2009, only two periods of heavy rainfall occurred, one in mid-June and the other in mid-July. The wind, temperature and humidity were similar for both years. A similar pattern was observed at Stora Bärby. Examination of the surrounding areas revealed no barberry plants. The calculated rate of disease progress (apparent infection rate), r, was between 0·26 and 0·55 when reported as incidence (proportion of stems with disease) and 0·17 and 0·38 when reported as severity (proportion of stem area covered with pustules) (Table 3). Figure 2 shows the disease progress at Fransåker during 2008 and 2009 and at Stora Bärby in 2009.

Table 3. Disease progress data for stem rust caused by Puccinia graminis f. sp. avenae in three oat fields in Sweden
LocationaYearNo. straws examinedDay of 1st infection r Incidenceb r Severityc
  1. aThe fields were scored at least three times after first visible symptoms and the number of straws examined was the same at all time points.

  2. bIncidence: number of stems with disease.

  3. cSeverity: proportion of stem area covered with pustules.

Fransåker2008401 August0·380·27
Fransåker20095026 July0·460·17
Stora Bärby20095010 July0·550·38
Figure 2.

 Disease progress curves reported as severity and incidence of stem rust caused by Puccinia graminis f. sp. avenae. The left legends show the average field scores of severity and incidence of disease (bars); the right legends and points shows the logit of the field scores. The trend line and equation show the calculated rate of disease progress.

Eleven of the 68 tested SSR markers, originally developed for P. graminis f. sp. tritici, successfully produced amplicons, indicating that they are useful for genotyping isolates of P. graminis f. sp. avenae.

In the main study, all loci were polymorphic, ranging from 6 to 28 alleles, and in total 139 alleles were found (Table 1). One locus, CAA53 was fixed in seven of the 17 populations. The observed heterozygosity (Ho) and expected heterozygosity (He) were similar for five of the 11 primer pairs. In the remaining six pairs, the observed heterozygosity was lower than the expected. Of the 11 loci, five had particularly large frequencies of null alleles (Pgestssr021, Pgestssr171, Pgestssr280, PgtCCA53 and PgtCAA93) (Table 1).

A total of 472 uredinia from 17 fields were successfully genotyped and used in this study and 457 unique genotypes were found. Barberry bushes were confirmed within a distance of 200 m from the surveyed field in Skarpenberga in 2008. In one field (Evertsholm in 2009) samples were taken on two occasions 3 weeks apart, as infected straws were only found in half the sampling sites within the matrix on the first occasion. In total, 44 samples were genotyped (15 from the first collection and 29 from the second). None of the samples within this field showed the same multilocus genotype (data not shown). Only the data from the second collection were used for further analyses, in order to retain a uniform sampling procedure for all populations (30 samples collected at one time point).

The genotypic diversity of the collected material was high, with G/N values of 0·90 or larger for all populations (Table 2). The number of observed alleles over all loci for each population ranged from 38 to 62. The loci Pgtssr109, Pgtssr279 and PgtCAA53 were all dominated by one allele (allele frequency 0·50 or higher), while the other loci were dominated by several or no alleles (Table S3). The observed (Ho) and expected (He) heterozygosity ranged from 0·330 to 0·501 and 0·449 to 0·618, respectively (Table 2). In general, the observed heterozygosity was lower than the expected for all populations except the one from Pattala, where the observed heterozygosity was higher than the expected. For three populations, those from Bettna, Bränne Övregård and Pattala, all from the year 2008, the FIS was close to zero, which indicates random mating and Hardy–Weinberg equilibrium (Table 2). All the other populations had FIS values between 0·130 and 0·487, indicating deviation from random mating. The selfing rate, s(g2), was significant for four field populations (Evertsholm, Pattala and Ultuna 2008, and Skarpenberga in 2009), with s(g2) values of 0·15, 0·19, 0·30 and 0·29 and P values of 0·023, 0·007, 0·001 and 0·001, respectively. The other field populations did not exhibit selfing rates different from zero (data not shown). In addition, the index of association (IA) ranged between −0·215 and 0·509 and was statistically significant only for the population collected at Klostergården Dala (Table 2). Thus, departure from linkage equilibrium could be shown only for this population.

The analysis of molecular variance (amova) showed that the majority of the variation (87%, < 0·001) was present within populations and only a minor part (13%) was caused by differences between populations (Table 4). amova also showed that variety and year only explained 2% (< 0·001) and 4% (< 0·001), respectively, of the differences between the varieties and between years (Table 4). The overall θ was 0·126 and the corresponding overall migration rate, Nm, was 1·7. Pairwise θ values ranged between 0·010 and 0·287 and were significant in all except the four cases when θ was 0·017 or lower (Table 5). Low and non-significant θ indicates no genetic difference between populations, while large values reflect differences between populations. The Mantel test performed for matrix correlation between genetic distance and the geographical distance showed no significant relationship (= 0·240).

Table 4. Analysis of molecular variance (amova) within and among Puccinia graminis f. sp avenae populations, collected from different oat varieties in 2008 and 2009, based on 11 microsatellite markers
Sourced.f.SSMSEst. var.% P-value
  1. aClone correction prior to each amova yielded different numbers of total observations.

Among populations16698·743·71·313<0·001
Within populations4503975·38·88·887
Totala4664674·0 10·1100
Among varieties496·224·10·22<0·001
Within varieties4113990·49·79·798
Totala4154086·6 9·9100
Among years197·297·20·44<0·001
Within years4384315·49·99·996
Totala4394412·6 10·3100
Table 5. Pairwise theta (θ; estimation of Wright’s FST) values under the diagonal for each population and their level of significance (< 0·001***, < 0·01** and < 0·05*) above the diagonal. Bold theta values are not significant
  1. aPopulations correspond to the locations on the map in Figure 1.

  2. b–fPopulations from the same farm are marked with the same letter: bFransåker, cIngvasta, dEvertsholm, eSkarpenberga, fBränne Övregård.

2008 1b ************************************************
 2c0·112 *********************************************
 3d0·1730·070 *******************************0·072******
 4e0·1530·0890·059 *************************************
 50·2580·1790·0760·081 ************************************
 6f0·1760·1560·1280·0840·110 *********************************
 70·1320·1300·1140·0890·1680·168 ******************************
 80·2190·1700·1040·0360·1010·1000·091 ***************************
 90·1800·0850·0210·0600·0660·0980·1440·090 **0·068******************
100·1810·1330·0540·0860·0470·1190·1350·1210·035 0·161****************
200911e0·1920·1150·0560·0700·0440·0980·1300·076 0·017 0·010  *****************
120·1810·1280·0870·1040·1010·1360·1470·1460·0540·0300·043 ***************
13b0·1700·1470·0930·1190·1600·1340·1390·1440·1200·1210·1200·080 0·103*********
14c0·1390·1510·0980·1140·1480·1040·1300·1200·0880·0980·0860·074 0·012  *********
15f0·1470·084 0·015 0·0460·0750·0990·1020·0920·0550·0570·0540·0920·0800·072 ******
160·2070·0900·0700·1080·1340·1720·1680·1710·0910·0810·1070·1050·1500·1660·093 ***


The genetic diversity of P. graminis f. sp. avenae causing stem rust on oats is large in Sweden. This, together with the presence of the alternate host, indicates that sexual reproduction is important for the epidemiology of this disease. The calculated rate of disease progress (r) was high, demonstrating the potential of the pathogen to cause severe damage and high yield losses (Table 3; Fig. 2). A high rate of disease increase means that the timing of the first infection is a crucial factor when deciding about fungicide treatment, although other factors, such as host-plant resistance and the weather, will affect the rate of disease development and the subsequent decision. Inclusion of the last data point for incidence data (close to 100%) for the field at Fransåker in 2009 could be debated, but rates calculated without this point were either very similar to or much larger than the ones presented here.

In most fields, the number of genotypes was equal to the number of samples, which indicates that a large number of individuals must have initiated disease in each field. It is likely that these initial spores arrived early and clustered in time, as the spores arriving early in a particular field are more important for an epidemic than the spores arriving late, according to O’Hara & Brown (1996). The possibility of overwintering on both wild and domestic oats is restricted because of the limited survival of Avena spp. in Sweden during the winter. The early arrival of a large number of genotypes may also limit the ability to detect the spread of clones between fields. The average temperature in Sweden during the growing season is slightly lower than the optimum of approximately 20°C for stem rust development (Stubbs et al., 1986). A reduced rate of disease development would also favour multiple, early infections. These could be initiated by aeciospores or by urediniospores of different genotypes from other oat fields. If only a few spores initiated the epidemic, several generations would be required for inoculum build-up and thus the initial signs of disease would appear only after several weeks, and the population in a field would be more clonal.

Simple sequence repeat primers developed for Puccinia graminis f. sp. tritici were useful in studying P. graminis f. sp. avenae. Four of the 11 primers gave a relatively high frequency of null alleles, which may reduce their usefulness for this particular pathosystem. The high null allele frequency of the selected SSR markers could be caused by the phylogenetic distance between P. graminis f. sp. avenae and P. graminis f. sp. tritici. Szabo (2007) reported that null alleles were common among the tested P. graminis f. sp. tritici isolates. Large numbers of null alleles are expected in species with large effective population sizes (Chapuis & Estoup, 2007). Null alleles may also be caused by contaminated samples, low quality DNA or failure in the PCR reaction. As the gene flow was relatively low (overall Nm value 1·7), null alleles could affect the results and the FST values (Chapuis & Estoup, 2007). A mixed sample consisting of two homozygotic individuals could also lead to an excess of heterozygotes within a particular locus, but this was not observed. The data set was analysed both with and without the EM correction algorithm (Chapuis & Estoup, 2007) and the results were similar. It is therefore concluded that the presence of the null alleles did not affect the markers’ usefulness.

The amova showed that most of the genetic variation was present within the different fields, and the variety and year of collection only affected the population structure to a small extent (Table 4). Both allelic and genotypic diversity were high in all fields. The positive values of FIS were consistent with random mating and Hardy–Weinberg equilibrium could not be rejected. The hypothesis that P. graminis f. sp avenae undergoes sexual reproduction in Sweden was also supported by the lack of associations among loci, as the index of association was statistically significant only for one of the 17 field populations. Inbreeding is probably responsible for the fact that the observed heterozyosity was lower than the expected heterogozity for some of the field populations as well as the significant selfing values. The basidiospores only travel a couple of hundred metres and the source of inoculum for infection on barberry must therefore be fairly close to a bush, whereas the aeciospores may travel hundreds of kilometres, just like urediniospores (Roelfs, 1985). Thus, somewhat locally distinct populations will develop and mating is not completely random, as barberry bushes play a role in the founding of local populations.

At the same time, some migration between geographical areas must occur, because neither the Mantel test nor the pairwise θ (Table 5) revealed any correlation between geographic and genetic distance and similar populations were recovered from fields that were separated by large geographical distances (the distance from Götala to Skarpenberga is approximately 400 km). Colonization of a field by a large number of different individuals, either from local barberry or through long distance dispersal, is followed by local subpopulation expansion. The rust populations studied here are diverse, but only a fraction of the rust in one area infects the plants in a field as the same locations in different years can produce different populations (Evertsholm pairwise θ between years = 0·210), or similar populations (Skarpenberga pairwise θ between years = 0·070).

A large influx of urediniospores with different genotypes could also produce similar results. Although P. graminis can spread by air over long distances (Kolmer, 2005) via the Western or Eastern European Tract, respectively, and spread to Scandinavia (Nagarajan & Singh, 1990), the absence of oats in Sweden’s neighbouring countries would reduce the importance of inoculum from outside the country. Therefore, long-distance dispersal is probably not as important for stem rust epidemics as is the spread of P. graminis f. sp. avenae within the country.

A population with several clonal lineages of P. graminis f. sp. avenae was described in an Australian study which also used microsatellite markers (Keiper et al., 2006). In this region, the alternate host is absent (the population must survive clonally) and this will lead to a different population structure. Sexual reproduction of Puccinia spp. as a result of the presence of an alternate host within an area will produce an abundance of spores with different genotypes, which can then initiate disease. The offspring could differ in aggressiveness and might have different ecological strategies in their mode of growth and reproduction, as well as if grown in competition with other individuals (Newton et al., 1999). When the fungus reproduces clonally, the different genotypes will specialize and adapt, and aggressive clones will probably take over and dominate the population (Pariaud et al., 2009). Reliance of the pathogen on survival as teliospores would also select for this particular trait, which Roelfs (1982) also observed for P. graminis in the northern United Sates. Eradication of barberry would not only decrease the number of genotypes but also delay the first occurrence of disease in the field, especially for a crop such as oats, which lacks a green bridge for overwintering (Roelfs, 1982).

The race diversity of P. graminis f. sp. avenae is strongly connected with disease development and the resistance of the host plant. Currently, it is not known which resistance genes are present in the commercially grown oat cultivars in Sweden. This study did not investigate which races of P. graminis f. sp. avenae are present in Sweden, but it can only be speculated whether the large number of genotypes found comports with the number of races that might be detected.

The allelic and genotypic diversity of P. graminis f. sp. avenae within and between Swedish oat fields is large. The disease progress is rapid, and the potential for yield loss is high even if the pathogen arrives in the fields late in the season. The alternate host, barberry, is present and the number of bushes has increased since the repeal of the barberry eradication law in 1994. It is highly likely that sexual reproduction of P. graminis occurs in Sweden, thus fitting the definition of a high-risk pathogen (McDonald & Linde, 2002). While fungicide application may be one disease control strategy, timing of application is very important. A possible method for reducing the amount of stem rust disease in the future and the risk of developing new races of the pathogen would be to re-initiate the removal of barberry in Sweden on a national scale.


The authors wish to thank Sam Stoxen, Jerry Johnson and Kim Nguyen at the Cereal Disease Laboratory for help with the method. This research was funded by the Swedish Farmers’ Foundation for Agricultural Research (SLF) and the Swedish University of Agricultural Sciences (SLU).