A novel form of resistance in rice to the angiosperm parasite Striga hermonthica


Author for correspondence: Julie Scholes Tel: +44 114 2224780 Fax: +44 114 2220002 Email: J.Scholes@Sheffield.ac.uk


  • • The root hemiparasitic weed Striga hermonthica is a serious constraint to grain production of economically important cereals in sub-Saharan Africa. Breeding for parasite resistance in cereals is widely recognized as the most sustainable form of long-term control; however, advances have been limited owing to a lack of cereal germplasm demonstrating postattachment resistance to Striga.
  • • Here, we identify a cultivar of rice (Nipponbare) that exhibits strong postattachment resistance to S. hermonthica; the parasite penetrates the host root cortex but does not form parasite–host xylem–xylem connections.
  • • In order to identify the genomic regions contributing to this resistance, a mapping population of backcross inbred lines between the resistant (Nipponbare) and susceptible (Kasalath) parents were evaluated for resistance to S. hermonthica.
  • • Composite interval mapping located seven putative quantitative trait loci (QTL) explaining 31% of the overall phenotypic variance; a second, independent, screen confirmed four of these QTL. Relative to the parental lines, allelic substitutions at these QTL altered the phenotype by at least 0.5 of a phenotypic standard deviation. Thus, they should be regarded as major genes and are likely to be useful in breeding programmes to enhance host resistance.


Parasitic angiosperms are a taxonomically and geographically diverse group of plants that comprise approx. 1% of the angiosperm flora (c. 4000 species). Parasitic plants attach to the roots or shoots of their host by a specialized organ, the haustorium, which abstracts water, nutrients and organic solutes from the host xylem and/or phloem. Striga hermonthica and S. asiatica are obligate root hemi-parasites that infect C3 and C4 grasses, most commonly maize, sorghum, millet and upland rice. They have been estimated to infest some 40% of the cereal-producing areas of sub-Saharan Africa; S. hermonthica alone may now infest over 10 million hectares and those most severely affected are subsistence farmers. Infestations of these crops result in severe grain losses, estimated at US$7 billion annually (Berner et al., 1995).

One of the reasons that Striga has such devastating impacts on the growth and yield of cereals relates to its dual mode of action. First, Striga plants compete effectively with the host for carbon, nitrogen and inorganic solutes (Frost et al., 1997; Gurney et al., 1999). Second, the parasite has a so-called ‘phytotoxic’ effect on the host plant within days of attachment (Berner et al., 1995; Musselman & Press, 1995; Frost et al., 1997; Gurney et al., 1999); a very small parasite biomass, with attachments of less than 4 mm in size, results in a large reduction in host height, biomass and eventually grain yield (Gurney et al., 1999). The mechanism underlying the ‘phytotoxic’ effect of Striga on the host has not yet been elucidated, but may involve the production of a ‘toxin’ (Musselman & Press, 1995). The lifecycle of Striga is intimately associated with that of its host; Striga seeds will only germinate in response to specific chemical cues (sesquiterpene lactones) that are present in root exudates (Yoder, 1999). Following germination, a sticky radicle attaches to the root of the host and, following perception of host-derived haustorial initiation factors, parasite cells invade the host cortex, reaching the host vasculature within a period of approx. 5 d (Albrecht et al., 1999). At this stage they form direct contact with host xylem vessels.

Both the life cycle of the parasite and its mode of action make it difficult to control: an adult plant can produce up to 100 000 tiny seeds that can survive in the soil for 20 yr or more. The ‘phytotoxic’ effect of Striga means that whilst control measures that lower numbers of emerged Striga plants may be effective in reducing the density of seeds in the soil seed bank (in the medium- to long-term), they are unlikely to have any short-term impact on crop yield. At present, perhaps the most effective strategies centre on agronomic practices, such as improving soil fertility (Cechin & Press, 1993a,b; Showemimo et al., 2002), or intercropping cereals with the legume Desmodium uncinatum (Khan et al. 2002). Some tolerant varieties of sorghum and maize (i.e. those that yield a little better in the presence of the parasite) have been identified (Gurney et al., 2002; Oswald & Ransom, 2004). Many of these tolerant varieties produce lower amounts of germination stimulants in their root exudates, leading to smaller numbers of attached parasites and/or to later attachment of the parasites to the host. However, because parasite biomass and loss of crop yield are not linearly related, tolerant varieties can still exhibit significant reductions in yield (Gurney et al., 1999). Breeding for postattachment resistance in cereals is likely to be the most cost effective and sustainable form of long-term control. Despite considerable efforts, by large numbers of researchers, to screen cultivars of sorghum and maize, and wild relatives of these species, for postattachment resistance to Striga, success has been limited. At present, no postattachment resistance to Striga has been found in cultivars of maize (Oswald & Ransom, 2004). Although parasites fail to develop on Tripsacum dactyloides, a wild relative of maize, the latter is difficult to use in traditional breeding programmes (Gurney et al., 2003). Recently, two sorghum cultivars (Framida and Dobbs) and a wild sorghum accession (P47121) have been shown to exhibit a hypersensitive-like necrosis at the site of attachment of S. asiatica parasites, although the molecular genetic basis of this resistance is unknown (Mohamed et al., 2003).

In this article, we report the discovery of a cultivar of rice that exhibits very high levels of postattachment resistance to S. hermonthica. In this case, the parasite penetrates the host root cortex but is unable to form vascular continuity with the host and it dies. To our knowledge, this is the first report of such resistance in any cereal host to this devastating parasite. Specifically, we describe the phenotype of the resistance response and, using a mapping population of backcross inbred lines, we identify seven major quantitative trait loci (QTL) that together explain 31% of the phenotypic variation in host resistance.

Materials and Methods

Plant material and mapping population

A large number of rice cultivars (Oryza sativa subspecies japonica and indica, O. glaberrima and wild relatives of rice) were screened for postattachment resistance to S. hermonthica. The cultivars screened are listed in Table 1, together with details of the source of the seed and their country of origin. Three cultivars were selected for detailed analysis: IAC 165 (very susceptible), Kasalath (susceptible/tolerant) and Nipponbare (resistant). Nipponbare is a japonica subspecies grown in lowland areas of Japan, whereas Kasalath is an indica subspecies that originated from Assam in India and is more characteristically grown in upland areas. Ninety-eight backcross inbred lines (BILs), developed from self-pollinating BC1F1 (Nipponbare/Kasalath//Nipponbare) plants for 14 generations by the single-seed descent method, were used for QTL analysis. A detailed description of the mapping population has been published previously (Lin et al., 1998). The seeds of Shermonthica used in this study were collected from plants parasitizing maize in Kibos, western Kenya, in 1997. S. hermonthica is an obligate out-crossing species, thus populations will be genetically variable.

Table 1.  Cultivated and wild species of rice screened for susceptibility to Striga hermonthica
Rice speciesVariety name (accession number)Seed sourceCountry of originSusceptibility to Striga hermonthica
  1. S, susceptible; R, resistant.

  2. LA, Long Ashton Research Station, Bristol, UK (now closed); IRD, Institute for Research and Development, Montpellier, France; IRRI, International Rice Research Institute, Philippines; RGRC, Rice Genome Research Centre, Japan.

  3. aStriga hermonthica seed was collected in Kibos, Kenya in 1990.

Cultivated rice
 Oryza sativa (japonica)IAC 165 (39045)IRRIBrazilS
 Oryza sativa (japonica)IAC 165LABrazilS
 Oryza sativa (japonica)CocodrieUSAS
 Oryza sativa (japonica)NipponbareIRRIJapanR
 Oryza sativa (japonica)NipponbareRGRCJapanR
 Oryza sativa (japonica)KoshihikariRGRCJapanS
 Oryza sativa (japonica)Azucena (328)IRRIPhilippinesS
 Oryza sativa (indica)IR64 (6690)IRRIPhilippinesS
 Oryza sativa (indica)IR64 (6697)IRRIPhilippinesS
 Oryza sativa (indica)KasalathRGRCIndiaS
 Oryza sativa (indica)AS2 (16320)IRRIIndiaS
 Oryza sativaIguape CatetoLAAfricaS
 Oryza sativaNamrooLAAfricaS
 Oryza glaberrima(G8-1)IRRIAfricaS
 Oryza glaberrima(CG14)IRRIAfricaS
 Oryza glaberrima(CG17)IRRIAfricaS
 Oryza glaberrima(CG20)IRRIAfricaS
 Oryza glaberrima(M27)LAAfricaS
 Oryza glaberrima(T2)LAAfricaS
Wild rice
 Oryza australiensis(OA4)IRRIAustraliaS
 Oryza australiensis(OA6)IRRIAustraliaS
 Oryza malapazaensis(W-1159)IRDUnknownS
 Oryza officinalisPhilippines (80773)IRRIPhilippinesS
 Oryza officinalisIndia (80766)IRRIIndiaS
 Oryza punctata(W1590)IRDAfricaS
 Oryza punctataTanzaniaLATanzaniaS
 Oryza rufipogonAustralia (OR-7)IRRIAustraliaS
 Oryza rufipogonMalaysia W (593)IRRIMalaysiaS
 Oryza rufipogonSouth America (W 1196)IRRISouth AmericaS
 Oryza brachyantha(W 0654)IRRIAfricaS
 Oryza granulate80739IRRIMyanmarS

Growth and infection of rice cultivars

Rice seeds were germinated between two sheets of moistened glass-fibre filter paper (GF/A Whatmann; BDH, Poole, UK) supported by a block of moistened horticultural rockwool (Aquaculture, Sheffield, UK). After 7 d, a single rice seedling was transferred to a root observation chamber (rhizotron). A rhizotron consisted of a 14-cm Petri dish filled with rockwool, onto which a mesh was placed (100 µm polyester multi; Plastic Group, Birkenhead, UK). Roots of the rice seedling grew down the mesh, and openings at the top and bottom of the Petri dish allowed for shoot growth and drainage, respectively. (Root growth was not restricted in these rhizotrons during the experimental period.) A black plastic sleeve prevented light from reaching the roots. Rhizotrons were supplied with 100 mL of 40% Long Ashton solution, containing 1 mol m−3 ammonium nitrate (Hewitt, 1966), at 48-h intervals. Plants were grown in a controlled environment room operating with a 12-h photoperiod and a photon-flux density of 800 µmol quanta m−2 s−1 at plant height. Day : night temperatures were maintained as 27 : 20°C, and day : night relative humidity was maintained at 50 : 70%. Twenty-one days postsowing, each rice plant was inoculated with 5 mg of preconditioned, pregerminated S. hermonthica seed by carefully aligning the seeds along the roots (Gurney et al., 2003). S. hermonthica seeds were artificially germinated (using a 0.1 ppm solution of the artificial germination stimulant, GR-24) to ensure synchronous attachment of the parasite to rice roots. After inoculation, the rhizotrons were returned to the controlled environment room in a completely randomized design.

Phenotype of resistance/susceptibility in rice to S. hermonthica

In order to screen cultivars for resistance to S. hermonthica, plants were infected and, after ≈ 21 d, were designated as either susceptible (supporting multiple Striga attachments) or resistant (supporting little development of Striga following attachment of the parasite). A detailed study to compare the development of S. hermonthica on IAC 165 (very susceptible), Kasalath (susceptible/tolerant) and Nipponbare (resistant) was carried out by recording the stage of development of the parasite on each plant 3, 9, 15 and 21 d after inoculation (dai) using a stereomicroscope (SUZD 338, Former USSR). Developmental stages were defined as: S1, the S. hermonthica radicle had attached to the host root and swollen to form an immature haustorium, the seed coat remained intact; S2, leaf primordia had emerged from the seed coat; S3, S. hermonthica shoots had three to five scale leaf pairs; S4, S. hermonthica shoots had six or more leaf pairs; D, parasites attached, failed to develop and died. Photographs of S. hermonthica plants on rice roots were taken using an SLR camera (RICOH XR-X 3PF; Ricoh, Eschborn, Germany) mounted on the stereomicroscope (Leica WILD M10; Leica, Milton Keynes, UK). Eight replicate rhizotrons were established for each treatment.

In order to examine the extent of parasite development within the host root cortex, representative haustoria of S. hermonthica were dissected from host plants at 3, 9, 15 and 21 dai. Samples were fixed, embedded and attached to adhesive-treated microscope slides (polysine slides; SLS, Nottingham, UK), as described in Gurney et al. (2003). After the removal of paraffin, slides were stained with Safranin O [1% (w/v) in 30% (v/v) ethanol; 5 min] and Astra blue [0.5% (w/v) in 2% (w/v) tartaric acid; 10 s). Sections were dried on a hot plate at 45°C for 1 h and mounted with DePeX (BDH). Sections were observed using an Olympus BX51 microscope (Olympus Optical Ltd, London, UK) and photographed using a digital camera (Olympus DP11; Olympus Optical Ltd).

Phenotypic recording of BILs

The mapping population comprised 391 plants (mean c. 4.0 replicate plants per BIL). A confirmatory mapping population (see below) comprised an additional two plants from each of the 98 BILs. Every line has previously been genotyped at 245 restriction fragement length polymorphism (RFLP) markers that span all 12 rice chromosomes at a mean intermarker interval of 4.8 cm. Each of the 391 BIL plants (and the 196 confirmatory plants) was established in the presence of S. hermonthica, as described above. At 21 dai, plants were scored for S. hermonthica resistance, with host resistance being defined as the proportion of S. hermonthica seeds at stages S1 or D. Because resistance was scored as a proportion, it was arcsine transformed whereby:


X is the number of S. hermonthica attachments that failed to advance beyond stage S1 at 21 dai and n is the total number of S. hermonthica parasites attached to the host plant.

QTL analysis

Resistance QTL were mapped by composite interval mapping (CIM) (Zeng, 1994) implemented in QTL Cartographer (Basten et al., 2001). Individual plants were treated as data points, with BIL fitted as a factor in all analyses. CIM tests for the presence of a QTL while conditioning for the effects of other QTL that lie on other chromosomes or at a minimum user-defined distance from the test location. Here, CIM used model 6 of the ZmapQTL program of QTL cartographer with a walking speed of 2 cm, a maximum number of background parameters (i.e. other QTL) set to 20 and a window size of 20 cm. Background parameters were identified by forward–backward stepwise regression conducted in the QTL cartographer program SrmapQTL, using the default parameters. CIM yields a likelihood ratio test (LRT) statistic at 2-cm intervals. Genome-wide and nominal statistical significance were obtained by permutation testing (1000 permutations) (Churchill & Doerge, 1994). All QTL that were genome-wide significant at P < 0.10 are reported. Note that marker genotypes were only available for generation BC1F5, which means that heterozygous genotypes (4.6% of the total) had to be scored as unknown in the mapping population, making the measurement of dominance effects of QTL impossible.

Confirmatory screen of QTL

Any QTL mapping screen carries a risk of false assignment of QTL (Type I error) owing to the large number of statistical tests employed. Therefore, we took the unusual, but robust, step of conducting a confirmatory screen for QTL using two plants from each of the 98 BILs. Measurement of resistance and QTL analyses were exactly as described above. Lander & Kruglyak (1995) regard QTL as confirmed if they yield a test statistic that exceeds a genome-wide significance of P < 0.05 in the first scan, supported by a nominal significance of P < 0.01 in the second scan. Because we were attempting to confirm seven QTL, we used a Bonferroni corrected significance threshold of P = 0.01/7 = 0.0014 to confirm linkage. An additional test to determine whether the two screens provided compatible results was based on Keightley & Knott (1999). Two mapping experiments can be compared by measuring the correlation in test statistic at every location across the genome. However, a test statistic at one location (e.g. a QTL peak) is usually very highly correlated with the test statistic at adjacent locations (e.g. a few cm from the peak). This autocorrelation makes standard tests of statistical significance inappropriate. Keightley & Knott (1999) devised an elegant solution to the problem, whereby statistical significance is obtained by permutation. In each permutation, the linear array of each chromosome was treated as a circle. The two circles were then aligned by rotating one past the other, stopping at a random position. For each alignment (permutation), the correlation coefficient between genome-wide test statisitics was calculated, thereby generating a null distribution for the correlation coefficient between the two experiments. This approach generates an estimate of statistical significance while accounting for the autocorrelation observed between test statistics at linked locations. One-thousand permutations were employed.

Results and Discussion

Characteristics and phenotype of resistance to S. hermonthica

Rice cultivars revealed striking differences in their ability to support the growth and development of S. hermonthica. A large number of rice cultivars (O. sativa subspecies japonica and indica, O. glaberrima and wild relatives of rice) were screened initially for their postattachment resistance to S. hermonthica; all were found to be susceptible, to varying degrees, with the exception of one cultivar, Nipponbare (japonica-type), which exhibited strong resistance.

In a susceptible interaction (IAC 165 or Kasalath) the parasite haustorium attached to the root system within 1 d of placing pregerminated seeds onto the roots. The initial haustorium forms in response to haustorial-inducing factors present in host root exudates (Estabrook & Yoder, 1998; Yoder, 1999; 2001). By day 3, the endophyte had penetrated both the root cortex (a process that is also responsive to host-specific factors) and the root endodermis, and the first parasite–host xylem–xylem connections were just visible (Fig. 1a,i and Fig. 2). In order for the parasite to form xylem–xylem connections, the endophye must penetrate the host endodermis and make direct contact with the host stele (Yoder, 1997), although little is known about the host factors required for successful penetration of the endodermis. Once xylem–xylem connections had formed, the parasites developed rapidly; by day 9, leaf primordia were well developed and the haustorial tissues had begun to differentiate (Fig. 1a,ii). A cross-section through the parasite haustorium at 21 dai revealed well-developed parasite–host xylem–xylem connections and a clearly differentiated vascular core and hyaline body (Fig. 1a,iii). A mature haustorium is essential for parasite growth and survival. First, xylem–xylem connections allow the movement of solutes from host to parasite and, second, the hyaline body is thought to metabolise these solutes, or at least to regulate the supply of nutrients to the developing parasite. The tissues of the hyaline body demonstrate intense metabolic activity and many of the enzymes involved in N metabolism are located in this region (Taylor, 2001).

Figure 1.

(a) Transverse sections of embedded tissue of susceptible (Kasalath) (I–iii) and resistant (Nipponbare) (iv–vi) rice roots 3, 9 and 21 d after inoculation with Striga hermonthica. The scale bar represents 0.1 mm. En, endophyte (internal part of haustorium); Hc, host root cortex; He, host endodermis; Hx, host xylem; Hx–Px, host–parasite xylem continuity; Hy, hyaline body; P, parasite haustorium; and Px, parasite xylem vessels. (b) The root systems of susceptible (IAC 165 and Kasalath) and resistant (Nipponbare) rice cultivars 25 d after infection with the angiosperm parasite Striga hermonthica.

Figure 2.

Development of Striga hermonthica on the roots of IAC 165, Kasalath and Nipponbare 3, 9, 15 and 21 d after inoculation. Stages of development of S. hermonthica were defined as: S1, S. hermonthica attached, seed coat is intact and a tubercle is evident; S2, emergence of leaf primordia; S3, S. hermonthica shoots have three to five scale leaf pairs; S4, S. hermonthica shoots have six or more leaf pairs; Dead, S. hermonthica plants attached and died.

Although both IAC 165 and Kasalath supported Striga development, there were clear differences in the rate of growth of the parasites; those on IAC 165 developed more quickly and were larger than those on Kasalath (Figs 1b and 2). Twenty-one days after inoculation, 89% of parasites attached to IAC 165 had between three and six scale leaf pairs (stages S3 and S4) and fewer than 1% of parasites were dead (Fig. 2). By contrast, only 36% of parasites attached to Kasalath had reached developmental stages S3 or S4 at this time, and 11% of parasites were dead. The difference in growth rate and developmental stage of the parasites may be the result of differences in the ability of the hosts to supply nutrients to the parasites, but it is more likely to reflect genotypic differences in the susceptibility/resistance of the two cultivars, with Kasalath exhibiting some resistance to the parasite.

The early stages of parasite development on Nipponbare were similar to those on Kasalath or IAC 165; parasites attached to the host root system within 1 d of inoculation and, by day 3, the parasitic endophyte had successfully penetrated the host root cortex, thus demonstrating that host-specific factors necessary for early haustorial formation and successful penetration of the cortex were present in this cultivar. At this early stage, the external appearance of the parasites on Nipponbare resembled those on susceptible varieties (Fig. 1a,i and iv); there was no evidence of a hypersensitive reaction, as observed in the sorghum cultivar Framida following infection with S. asiatica (Mohamed et al., 2003). However, in contrast to Kasalath or IAC 165, the parasite rarely breached the endodermis in Nipponbare roots. The reason for this is unclear as there was no obvious difference between the endodermis of Nipponbare and that of susceptible cultivars; for example, it did not appear to be more heavily lignified (Fig. 1a). By 9 dai it was clear that the parasite was unable to form parasite–host xylem–xylem connections; in many cases, the endophyte passed straight through the root cortex and emerged from the other side of the root, or it encircled the vascular core of the host within the root cortex (Fig. 1a,iv–vi). Vascular continuity between host and parasite not only allows the transport of water and nutrients from host to parasite, but may also provide factors required for further differentiation of the haustorium. The haustorium of parasites attached to Nipponbare did not mature and differentiate (Fig. 1a,vi). Thus, by day 21, 49% of the attached parasites were either dead or showed signs of necrosis (Fig. 1a,vi inset, Fig. 1b and Fig. 2). Only 2% of attachments developed beyond the primordia leaf stage, and stem elongation was rare.

It is clear that Nipponbare is resistant to S. hermonthica, at least in part because the parasite failed to form vascular continuity with the host. This type of resistance differed from that observed in the sorghum cultivars Framida and Dobbs where, following attachment of S. hermonthica, a proportion of the parasites died owing to the onset of a rapid hypersensitive reaction (Mohamed et al., 2003) and also in the nonhost interaction between S. asiatica and marigold. In the latter, penetration of the cortex was terminated early and the endophyte rarely reached the endodermis (Hood et al. 1998). Gowda et al. (1999) used differential display techniques to identify host genes expressed in marigold roots in response to penetration by S. asiatica. These authors isolated and cloned a novel gene (NRSA-1) that had homology to known plant disease-resistance genes, but which differed structurally from those previously described. A heavy deposit of lignin was often also observed around the invading endophyte, and haustorial cells at the host–parasite interface showed signs of necrosis. Recently, resistance to S. hermonthica has also been identified in T. dactyloides, a wild relative of maize (Gurney et al., 2003). In this case, S. hermonthica arrested after host–parasite xylem–xylem continuity was established. The haustorium failed to differentiate either because the host lacked signals required for haustorial development or, more likely, because T. dactyloides produced a signal that inhibited haustorial development. The resistance observed in Nipponbare is most similar to that observed in the resistant interaction between vetch Vicia atropupurea cultivar Popany and the angiosperm parasite, Orobanche aegyptiaca. In this interaction, the parasite haustorium penetrated the root, but was blocked at the root endodermis layer (Goldwasser et al., 2000). However, in contrast to Nipponbare, the blockage was coupled with a large secretion of unknown composition, which prevented the parasite from establishing vascular continuity.

Nipponbare is resistant to S. hermonthica as the parasite fails to form xylem–xylem connections, thus preventing the transport of water, nutrients and developmental cues that allow successful development of the parasite. The most common resistance mechanisms are those in which the host lacks factors needed by the parasite, particularly germination stimulants (Parker & Riches, 1993), but this is the first example of resistance caused by a lack of formation of parasite–host vascular continuity. The majority of resistance/tolerance factors are not simply inherited, as illustrated by a recent study of the genomic regions influencing low germination stimulant production (and hence resistance/tolerance to S. hermonthica) in two recombinant inbred populations of sorghum (Haussmann et al., 2004). Such studies suggest that resistance is polygenic in nature and hence extremely important in an agricultural context. In order to begin to elucidate the genetic basis of resistance in Nipponbare, a QTL analysis was undertaken utilizing a mapping population of BILs (Nipponbare/Kasalath//Nipponbare).

The genetic basis of resistance to S. hermonthica

The distribution of S. hermonthica resistance in the mapping population is consistent with a polygenic mode of inheritance (Fig. 3). The mean resistance in the mapping population was intermediate between that observed in the parental lines, although the distribution of resistance in the mapping population did show slight bimodality, which may be consistent with some genes of major effect segregating. The range in resistance in the mapping population varied from more susceptible than the most susceptible Kasalath plant to more resistant than the Nipponbare parental line (Fig. 3).

Figure 3.

Distribution of resistance to Striga hermonthica in the mapping population, F1BC14 (grey bars), and in the parental lines Kasalath (white bars) and Nipponbare (black bars). n = 16 for parental lines; n = 391 for the mapping population.

Seven QTL were identified at the genome-wide P < 0.10 level (Table 2, Fig. 4), four of which provided test statistics that were never exceeded in 1000 permutations (genome-wide P < 0.001). Unsurprisingly, the Nipponbare allele conferred greater resistance than the Kasalath allele at six of the seven QTL. However, a QTL of large effect on chromosome 4 showed a contrasting effect; the Kasalath allele was more resistant than the Nipponbare allele. In practical terms this discovery is exciting; introgression of the Kasalath allele into a Nipponbare background would result in a phenotype even more resistant than Nipponbare. Consistent with this is the observation that a small number of BILs exhibited greater resistance to S. hermonthica than Nipponbare (Fig. 3).

Table 2.  Striga resistance quantitative trait loci (QTL) significant at the genome-wide threshold (P < 0.10)
ChromosomePositionLRTPPVEAllelic substitution (standard deviations)Confirmed
  • LRT, likelihood ratio test statistic where the null hypothesis is no QTL. P, nominal significance of the LRT determined by 1000 permutations of the data. PVE, percentage phenotypic variance in the mapping population explained by the QTL. The additive effect on mean resistance (arcsine transformed) of an allelic substitution from a Kasalath allele to a Nipponbare allele (the effect size is also measured in standard deviations, where the phenotypic standard deviation in the parental races is 0.08). An additive effect with a positive coefficient means that the Nipponbare-derived allele confers increased resistance. The final column describes whether the QTL was confirmed in the second screen. A QTL was confirmed if the second screen yielded a nominal statistical significance of P < 0.0014.

  • ***

    Genome-wide significant at P < 0.001;

  • **

    genome-wide significant at P < 0.05;

  • *

    genome-wide significant at P < 0.10.

146 cm17.74< 0.001*1.8 0.036 (0.45)No
479 cm66.13< 0.001***7.6−0.064 (−0.8)Yes, P < 0.0014
577 cm19.65< 0.001*1.9 0.039 (0.49)Yes, P < 0.0014
697 cm45.04< 0.001***4.2 0.051 (0.64)Yes, P < 0.0014
774 cm57.56< 0.001***5.5 0.060 (0.75)No
832 cm21.59< 0.001**2.1 0.038 (0.48)No, P = 0.004
1241 cm63.28< 0.001***7.4 0.075 (0.94)Yes, P < 0.0014
Figure 4.

Comparison of the estimates of genome-wide quantitative trait loci (QTL) additive effects for the initial (unbroken black line) and confirmatory (dotted line) screens. Estimates of QTL additive effects on host resistance to Striga hermonthica (arcsine transformed) are plotted for the entire rice genome (chromosomes 1–12), for the initial screen (black curve) and for the confirmatory screen (dotted curve). Additive effects of > 0 are consistent with the Nipponbare allele conferring resistance. The test statistics at each location are significantly positively correlated between screens (r = 0.660; P = 0.0012), indicating that the findings of the first screen are broadly confirmed by the findings of the second screen. Note that the apparent larger effect sizes observed in the confirmatory screen are probably attributable to an effect size overestimate caused by a relatively small sample size (n = 196). Vertical lines indicate a chromosome boundary. Positions of the seven QTL identified in the initial screen are shown. *Genome-wide significant at P < 0.1; **genome-wide significant at P < 0.05; ***genome-wide significant at P < 0.001.

The magnitude of the QTL effects varied from 1.8% to 7.6% of the variance in mapping population resistance, and between them accounted for 31% of the overall variance. An alternative, and for breeding purposes more relevant, method of measuring QTL magnitude is as an effect size relative to the phenotypic variance observed in the parental races (Orr, 2001). Here, resistance in both Nipponbare and Kasalath had a standard deviation of approx. 0.08 (Table 3), while the effects of an allelic substitution at the QTL ranged from 0.036 to 0.075 (Table 2). Thus, under the usual definition of an allelic substitution that alters phenotype by at least 0.5 of a phenotypic standard deviation, these QTL should be considered major genes (Falconer & Mackay, 1996). The QTL on chromosomes 4 and 12 had the largest effect on S. hermonthica resistance; allelic substitutions at these two QTL approached 1 standard deviation (SD), and between them they explained 15% of the variance in resistance.

Table 3.  Mean resistance of the mapping populations and the parental lines
PopulationnMean resistance (SD)
Kasalath 160.740 (0.080)
Nipponbare 161.188 (0.078)
Mapping population3910.986 (0.185)
Confirmatory population1961.053 (0.210)

We took the unusual step of performing a confirmatory genome scan for QTL. Four of the seven QTL were detected in the second scan (at nominal significance P < 0.0014), and can therefore be regarded as confirmed linkage (Lander & Kruglyak, 1995). The QTL on chromosomes 1, 7 and 8 were not confirmed. The genome-wide test statistics were highly and significantly correlated between mapping experiments (r = 0.660, P = 0.0012, Fig. 4), indicating that the results from the first experiment were largely replicated in the confirmatory screen. Note the striking similarity between the genome-wide estimates of QTL effect in the two screens (Fig. 4). At some QTL (chromosomes 4, 5, 8 and 12), the confirmatory screen effect size was larger than that reported for the original screen. However, we report effect sizes as those estimated from the first screen because effect sizes may be upwardly biased in the smaller confirmatory screen (Beavis, 1994).

A key challenge in the future is to fine map, and ultimately to identify, the causative mutations or quantitative trait nucleotides (QTN) that are responsible for the S. hermonthica QTL. We aim to use an integrated microarray and fine-mapping approach to achieve these goals. Genes that show differential expression in infected tissues of the two parental races, and that map closely to the QTL, are excellent functional candidates for host-resistance genes (Beavis, 1994; Schadt et al., 2003). Fine mapping of the QTL can be achieved by the creation of nearly isogenic lines (NILs) and the subsequent scoring of resistance within those lines. Recent genetic tools for indica and japonica rice varieties include a large database of single nucleotide polymorphisms (Feltus et al., 2004), first draft genome sequences (Goff et al., 2002; Yu et al., 2002) and microarray technology. Thus, future prospects for the discovery of S. hermonthica resistance QTN are encouraging. Furthermore, the high degree of conserved synteny exhibited between many plant species (Salse et al., 2002) means that QTL identified in this system may rapidly be investigated in the most economically important crops in sub-Saharan Africa.


We thank: The Rockefeller Foundation for financial support; Dr Andy Greenland (Crop Genetics Research, Syngenta, Jealott's Hill Research Station, Bracknell, Berkshire), for advice and comments on the manuscript; and Professor M. Yano and the Rice Genome Resource Centre, Japan, for supplying the Nipponbare/Kasalath BILs used in this study.