| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Plant Biotechnology Centre, Agriculture Victoria, La Trobe Univ., Bundoora VIC 3083, Australia
b School of Biological Sciences, Univ. of Wales, Bangor, Gwynedd LL57 2UW, UK
c CSIRO Plant Industry, Long Pocket Lab., 120 Meiers Road, Indooroopilly QLD 4068, Australia
d Int. Crops Research Inst. for the Semi-Arid Tropics, Patancheru AP 502 324, India
* Corresponding author (j.r.witcombe{at}bangor.ac.uk)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Abbreviations: BLUP, best-linear-unbiased-prediction • CIM, composite interval mapping • cM, centiMorgan • India/Ghs, glasshouse screen in India • LG, linkage group • QTL, quantitative trait locus • REML, restricted maximum likelihood • RFLP, restriction fragment length polymorphism • UK/Ghs, glasshouse screen in the UK
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Breeding for resistance is the only practical and economically feasible way to control downy mildew in pearl millet (Singh et al., 1997). To screen for resistance, a reliable downy mildew field screen has been developed that imitates the natural field situation (Williams et al., 1981). Inoculum is provided by oospores ploughed into the soil and airborne sporangia from preplanted rows of infected pearl millet genotypes. Glasshouse screens have also been developed, in which pot-grown seedlings at the coleoptile stage are spray-inoculated with a suspension of sporangia (Singh and Gopinath, 1985). Compared with field screens, a larger quantity of material may be screened in a smaller space and across a shorter period of time (Singh et al., 1997). However, despite key differences in the screening methodology, there have been no published data to indicate that disease scores in the glasshouse screen reflect those that would be obtained in the field. One major methodological difference is that the soilborne oospores used as inoculum in the field are absent from glasshouse screens. Oospores can play a significant role in infection, and genes controlling resistance to sporangial inoculum can not be assumed to be the same as those that confer resistance to oosporic inoculum (Hash et al., 1997).
The inheritance of resistance to downy mildew is a quantitative character, which has hindered the study of resistance as well as effective breeding. The advent of molecular markers has enabled QTLs for resistance to be mapped and has shown that the quantitative nature of resistance is due to the segregation of QTLs for pathotype-specific resistance of major effect, along with QTLs of minor effect (Jones et al., 1995). Breeders can now use marker-assisted selection to incorporate specific resistance QTL into parental lines of existing high yielding lines (Jones et al., 1995). To increase the durability of resistance, several gene pyramiding and deployment strategies have been proposed (Hash et al., 1997; Witcombe and Hash, 2000), all of which require the availability of a pool of QTLs with different pathogen specificities. These QTLs may be mapped using multilocational field evaluations against a range of S. graminicola pathogen populations, but heritabilities from such field evaluations are often low (Hash et al., 1997). Alternatively, evaluation against a multinational range of pathogen populations can take place under glasshouse conditions in a region where S. graminicola does not pose any threat to crop production (Hash et al., 1997). To provide the latter resource, a glasshouse screening center has been established in the UK. To determine the effectiveness of this center, it is important to test whether these glasshouse results are informative relative to the field environment.
Quantitative trait locus analysis has been used to study the stability of fungal-resistance QTL across environments (e.g., Saghai Maroof et al., 1996). In a similar way, QTL analysis can be used to determine the validity of a glasshouse screen in relation to a field screen. In this case, the principal question is not whether QTLs are stable across environments, but are the QTLs detected in the field also detected under glasshouse conditions, and, are the direction and magnitude of QTL effects consistent? Here we test, by the use of QTL mapping, whether the glasshouse screens established in the UK and India are effective indicators of resistance in the field.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Insufficient seed of the P 7-3 resistant parent meant that for the disease screen in the UK, the sister line, P 7-4, was used in its place. In test screens, P 7-3 and P 7-4 were not significantly different in their resistance response to a number of pathogen populations (data not shown), suggesting that they are genetically similar in terms of their resistance. The susceptible 7042(S)-1 parent was used in all screens. The hybrid HB 3 (Dave, 1987), which was known to have a high degree of susceptibility to the ICRISAT-Patancheru pathogen population, was used as an additional susceptible check.
For the screens at ICRISAT-Patancheru, the pathogen was maintained in the field on the highly susceptible host-genotypes 7042(S) (an inbred line from which 7042(S)-1 was derived by selfing) and HB 3. In the UK, pathogen material was maintained on 7042(S).
Disease Screens
Field Screen
The field screen was carried out in the downy mildew field nursery (Singh et al., 1997; Williams et al., 1981) at ICRISAT-Patancheru, India. Approximately 20 seeds of each entry were sown in alfisol soil in 2-m-long rows, 0.75 m apart. Inoculum was provided by soilborne oospores, the levels of which had been built up over the years by ploughing infected plants into the soil following each screen, and airborne sporangia from rows of infected 7042(S) and HB 3 genotypes planted every 9th row. Entries were randomized within each of six replicates. Parents and checks were represented twice within each replicate. Plants were drip-irrigated daily. Disease incidence (percentage of plants showing disease symptoms) was assessed 21 (India/Field-1) and 40 (India/Field-2) d after sowing. Symptoms were seen as distinct chlorosis of infected leaves and, at 40 d after sowing, green ears, where panicle seed was replaced by leaf-like structures (phylloidy).
Glasshouse Screen in India
The glasshouse screen in India (India/Ghs) was carried out at ICRISAT-Patancheru, India. Between 40 and 60 seeds of each genotype were sown into a dry compost consisting of equal proportions of alfisol, farmyard manure, and fine sand. Each pot represented a replicate of a pearl millet genotype, and there were three pot replicates per genotype. Pots were placed in a completely randomized design in a glasshouse, where cool air blowers were initiated when the temperature exceeded 25°C. Irrigation was performed daily with an overhead hose. When the seedlings were at the coleoptile to one-leaf stage the inoculum was prepared. Infected leaves from mature 7042(S) and HB 3 plants growing in the downy mildew field nursery at ICRISAT-Patancheru were detached, wiped clean of existing sporangiophores using moist cotton wool and incubated in darkness in plastic boxes for 8 h at 20°C and 100% relative humidity. The resulting sporangia were harvested into tap water at room temperature and the concentration was assessed and adjusted to 
1.5 x 105 sporangia mL-1. Each pot of seedlings was sprayed with 5 mL of inoculum using a hand-pumped sprayer. Disease incidence (percentage of plants showing chlorotic leaf symptoms) was assessed 14 d later, when the plants had four to five leaves.
Glasshouse Screen in the United Kingdom
The glasshouse screen in the UK (UK/Ghs) was carried out at Pen-y-Fridd Experimental Station, Bangor, UK, according to the protocols described by Jones et al. (1995). The UK screen differed from the Indian glasshouse screen as follows: between 30 and 40 seeds of each pearl millet genotype were sown into a low-nutrient peat and sharp sand compost (Chempak Seed Base, Chempak Products, Hoddesdon Herts., UK; NPK 25-39-30 mg L-1). Four pot-replicates of each entry were flood-irrigated daily in a glasshouse providing a 16-h daylength (0600–2200 h) with supplementary lighting (to give a light intensity of between 500 and 1200 µmol m-2 s-1) and a temperature of 25 to 30°C from 0600 to 1800 h and 20°C from 1800 to 0600h. For the inoculum, pearl millet leaves containing abundant oospores were collected from diseased 7042(S) and HB 3 plants growing in the downy mildew field nursery at ICRISAT-Patancheru 2 yr prior to the screen. This leaf material was dried, ground, and stored at room temperature. In the UK, seed of 7042(S) was sown into compost mixed with the ground leaf material. Sporangia produced on infected plants were used to infect further 7042(S) plants to supply sufficient inoculum for the screens (Jones et al., 1995). The sporangia produced following incubation of infected leaves were harvested into chilled (<2°C) distilled water and the concentration was adjusted to 9 x 104 sporangia mL-1. Each pot of seedlings was sprayed with 4 mL of inoculum using a compressed-air cylinder-fed sprayer. The inoculum was maintained on ice to delay zoospore release until after inoculation and to ensure a uniform sporangial concentration with time (Jones et al., 2001).
Restriction Fragment Length Polymorphism Analysis and Map Construction
Restriction fragment length polymorphisms were detected with low copy probes derived from a PstI-restricted genomic library of the pearl millet genotype 7042(S)-11, sister line to 7042(S)-1 (Liu et al., 1994). Southern hybridization analysis was performed as described by Liu et al. (1994). Probes detecting polymorphism were hybridized to the restricted DNA of 125 F2 progeny. Some RFLPs contained a low to moderate frequency of nonparental alleles. Such RFLPs caused problems when ordering markers during map construction and were therefore removed from the mapping data set.
Linkage analysis was carried out using MAPMAKER 3.0 (Lander et al., 1987). Markers were grouped at LOD 4.0 and marker orders determined at LOD > 2.0. Orders were tested using the ripple command. Map distances were calculated using the Haldane mapping function (Haldane, 1919). Eighteen markers common to the pearl millet consensus map (Devos et al., 1995; Liu et al., 1994) were used to designate and orientate linkage groups (LGs).
Statistical Analysis of Disease Data
Disease scores made on genotypes with less than eight plants per pot-replicate were removed to avoid inaccurate assessment of percentage disease incidence. Data analyses were carried out in SAS (SAS Institute, 1985). Data were transformed (square root) to improve the distribution of residual-fitted data. Each screen was analyzed separately using general linear modeling (Proc GLM). Broad sense heritabilities were calculated for F4 family means according to Wricke and Weber (1986). All screens were analyzed together using restricted maximum likelihood (REML; Proc MIXED). F4 families and screens (i.e., environments) were treated as random effects.
The Shapiro-Wilk statistic (W-test) was used to assess the normality of the disease distributions for F4 family means (normal option in Proc UNIVARIATE). Spearman's rank-order correlation coefficients were determined for pair-wise comparisons of screens (spearman option in Proc CORR).
Quantitative Trait Locus Analysis
QTL analysis was carried out using CIM (Jansen, 1993; Zeng, 1993) in QTL/Cartographer 1.14 (Basten et al., 1994, 2000). One hundred and fourteen lines had both genotypic and phenotypic information, and 38 mapped markers were used. Marker cofactors were selected by forward–backward regression with a critical P-value of 0.10. A window size of 10 centiMorgans (cM) and a maximum of 10 marker cofactors per model were used. Threshold levels were estimated using P < 0.05 for the comparison of 1000 data permutations without re-selection of background markers (Churchill and Doerge, 1994). For each screen, QTL analysis was carried out using nontransformed F4 family means, square-root-transformed means, and predicted means [derived using best-linear-unbiased-predictions (BLUP) of means adjusted for replicate using REML]. To determine average effects across all screens, across-screen predicted means were derived using BLUP within REML. Multiple interval mapping (Jiang and Zeng, 1995) in QTL/Cartographer 1.14 (Basten et al., 1994, 2000) was used to carry out joint analysis of all screens and to test for QTL x screen (environment) interactions.
For each screen, the presence of digenic interactions were investigated using two-way analyses of variance (proc GLM) using all possible pair-wise marker combinations. Significant (P < 0.01) interactions were placed in a regression model with markers having a significant main effect to determine whether the model fit was improved.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
|
Phenotypic correlations between screens ranged between 0.51 and 0.75, and were all highly significant (Table 2). The highest correlation was for resistance assessed in India/Field-2 and UK/Ghs. When all screens were analyzed together using REML, the F4 family and F4 family x screen interaction variance components were highly significant (values were substantially larger than their standard errors; Table 3).
|
|
Quantitative Trait Locus Analysis
Comparison of Data Transformations
Square-root transformations of F4 family means within each screen resulted in an increase in the normality of the distributions, but did not affect the LOD profiles (data not shown). Predicted (BLUP) means were more normally distributed (only India/Field-1 still significantly deviated from normality) and resulted in quite different LOD profiles for the more skewed traits, India/Field-1 and India/Ghs. QTLs were detected in the same locations, but the variances explained by each QTL were considerably altered, with values becoming more similar to those of the less skewed traits, India/Field-2 and UK/Ghs. The analysis presented here is based on predicted means.
Comparison of Quantitative Trait Loci Detected in Each Screen
Two QTLs were detected in the same locations on LGs 1 and 2 in both assessments in the field (Table 4; Fig. 2A)
. The same two QTLs were also detected in the glasshouse screens carried out in India and in the UK (Table 4, Fig. 2B). The QTL on LG1 explained the highest proportion of the phenotypic variation in each of the screens (Table 4). For India/Field-2, India/Ghs, and UK/Ghs this was between 50 and 60% of the total phenotypic variation, whereas for India/Fld-1 the variation explained was about half of this at 31%. The QTL on LG2 explained between 10 and 16% of the total phenotypic variation. At each of the two QTLs detected in each of the screens, resistance was always inherited from the resistant parent (Table 4). The estimated inheritance of resistance varied between being partially dominant to completely dominant for the QTL on LG1, and between being additive and recessive for the QTL on LG2 (Table 4).
|
|
Analysis with multiple interval mapping detected the same two QTLs plus an additional two QTLs on LGs 3 and 5 (Fig. 2C). The locus on LG3 had been apparent just below the threshold level for QTL detection for the screen India/Field-2 (Fig. 2A). All QTLs had a significant QTL x environment component (data not shown).
Digenic Interactions
The number of pair-wise marker interactions that were significant at P < 0.01 was three times greater than that expected by chance. In each screen, and for the across-screen predicted means, the most significant of these occurred between M248 on LG3 and M390 on LG5. The same dominant x additive and dominant x dominant interactions were consistently observed (data not shown). These markers were in the same locations as the additional QTL detected with multiple interval mapping. An additional interaction of smaller effect was significant for India/Field-1, between the same marker, M248 on LG3, and M515 on LG1. Here, interactions were dominant x additive and additive x additive.
The interaction terms improved the significance of the overall model for each screen and increased the total phenotypic variation explained by up to 10% (Table 5).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The consistency of the resistance phenotype was further demonstrated by highly significant correlations between data collected in different screens. The highest correlation coefficient was found for comparisons between India/Field-2 and UK/Ghs, despite these screens being the least similar in terms of screening methodology.
Minor differences between the screens that could have affected the expression of QTLs included soil type, day and night temperatures, and irrigation regimes. Major differences were the type and method of inoculation, plant age at assessment, and the virulence composition of the inoculum. Several studies have shown that different inoculation techniques can result in different QTLs being detected (Albar et al., 1998; Lübberstedt et al., 1999; Setiawan et al., 2000; Zhu et al., 1999). For example, in sugar beet, out of eight QTLs detected in field and leaf-disc screens for resistance to Cercospora beticola, only one was common to both screens (Setiawan et al., 2000). Many studies have also found that plant age at inoculation and disease assessment result in different QTLs being detected (Clements et al., 2000; Qi et al., 1999; Shäfer-Pregl et al., 1999; Steffenson et al., 1996; Welz et al., 1999). Here, in the field, inoculum was in the form of soilborne oospores and airborne sporangia, and plants were exposed to inoculum from seedling emergence to grain filling. In the glasshouse, inoculum was solely in the form of a spray-inoculated suspension of sporangia at the coleoptile-1 leaf stage and disease assessment was carried out at the 4-5 leaf stage, long before flowering was initiated. Sporangia and oospores have different modes of infection, with sporangia producing zoospores that penetrate leaf epidermal cells and stomata (Subramanya et al., 1983; Mauch-Mani et al., 1989) and oospores infecting the meristematic region of the root (Safeeulla, 1976). The relative roles of oospores and sporangia of S. graminicola on disease progression where both are present is not clear, but probably depends on environmental conditions (Jeger et al., 1998). In this study, if oospores are assumed to be infecting plants in the field, the detection of the same QTLs in field and glasshouse screens suggests that oospores incur a similar response to infection as the zoospores that are released from sporangia. Similarly, these results suggest that inoculation at the coleoptile-1 leaf stage elicits a similar disease response as inoculation via continual exposure as the plant matures.
A major factor that could have affected the disease response was the virulence composition of the inoculum used within the different screens. For the screen in the UK, oospores were collected from the field 2 yr before inoculum was collected for the screens carried out in India. The virulence structure of pathogen populations of S. graminicola can rapidly shift at a single location (Singh and Singh, 1987; Thakur and Rao, 1996), but the pathogen components affecting resistance in this cross appear to have remained stable across these 2 yr. Changes in host-pathogen interactions due to shifts in inoculum pathogenicity is an additional factor that has to be considered in repeated disease screens, and may explain inconsistencies in QTL detection that have occurred in other resistance QTL studies.
One explanation for the consistent detection of QTLs across different environments in this study is that resistance to downy mildew in pearl millet has an unusually high heritability for a quantitative disease trait. This, in turn, is probably due to resistance being controlled by a small number of QTLs which are likely to be race-specific (Jones et al., 1995). Inconsistencies have often been found in other host-pathogen systems where many QTLs of small and medium effect were segregating, so that screen-specific QTLs were more likely to occur. Here, the consistent detection of a digenic interaction across screens was more surprising, as the effect of interacting QTLs may be expected to be more environmentally variable. Consistent detection of interactions across screens has rarely been reported, but was found by Saghai Maroof et al. (1996) for gray leaf spot in maize, where consistency was attributed to adequate disease pressures across screens.
Two methods for analyzing data across all screens were tested. Analysis of QTLs using across-screen predicted means resulted in the same two QTLs of main effect being detected that had been found in individual screens. Multiple trait analysis using CIM resulted in an additional two QTLs being detected, which were in the same location as the two regions that were found to be involved in digenic interactions in individual screens. As has been predicted (Jiang and Zeng, 1995; Korol et al., 1995), multiple trait analysis using correlated trait complexes appears to increase the power of QTL detection compared with single trait analysis.
Quantitative trait locus x environment (QTL x E) interactions were highly significant at each of the QTLs detected. However, the sign of the additive effect did not change between screens demonstrating that the QTL x E interactions observed were due to changes in magnitude, rather than direction, of QTL effects. The change in magnitude of the phenotypic variance explained by QTL between screens appeared to be associated with the normality of the disease distribution, rather than any differences in screening methodology: Improving the normality of the distribution for the more skewed data greatly affected the variance explained by detected QTL.
One aim of this study was to detect new resistance QTL for marker-assisted selection. However, the QTL with the largest effect on LG1 mapped to the same location as a QTL of major effect detected against this same pathogen population in a different pearl millet cross (Jones et al., 1995), suggesting that the QTLs are due to effects at the same resistance gene. Although the resistance sources for each cross were derived from different African germplasm collections, both had origins in the highly genetically diverse Souna landrace (Singh et al., 1990; Talukdar et al., 1998). Both host genotypes were selected at ICRISAT-Patancheru during the late 1970s and early 1980s, so that the same QTL could have been selected. The QTL on LG2 has not previously been mapped, but the apparent recessive inheritance of resistance at this QTL would render its use impractical in hybrid breeding. Equally, the two QTL that were only detected following analysis of pair-wise interactions are unlikely to be used. The possible incomplete genome coverage provided by RFLP in this study may mean that additional QTLs remain undetected, although the relatively high total phenotypic variation already explained (up to 64%) would suggest that any additional QTL do not have a major effect.
Further QTLs need to be detected for resistance gene-deployment or pyramiding strategies to be effective against downy mildew in pearl millet. This will require methodical screening of multiple mapping populations against a range of pathogen populations of S. graminicola under controlled conditions. We have shown that such screens carried out in the glasshouse in the UK and India are likely to result in the detection of QTLs that control resistance under field conditions.
|
ACKNOWLEDGMENTS |
|---|
Received for publication July 13, 2001.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
