Evolution of the wheat blast fungus through functional losses in a host specificity determinant

See allHide authors and affiliations

Science  07 Jul 2017:
Vol. 357, Issue 6346, pp. 80-83
DOI: 10.1126/science.aam9654

Genetic analysis of disease emergence

In the 1980s, wheat crops began to fall to the fungal pathogen that causes blast disease. First seen in Brazil, wheat blast last year caused devastating crop losses in Bangladesh. Inoue et al. tracked down the shifting genetics that have allowed the emergence of this potentially global threat to wheat crops (see the Perspective by Maekawa and Schulze-Lefert). Wheat varieties with a disabled resistance gene were susceptible to pathogen strains that affected oat and ryegrass crops. Subsequent genetic changes in the pathogen amped up the virulence in wheat.

Science, this issue p. 80; see also p. 31


Wheat blast first emerged in Brazil in the mid-1980s and has recently caused heavy crop losses in Asia. Here we show how this devastating pathogen evolved in Brazil. Genetic analysis of host species determinants in the blast fungus resulted in the cloning of avirulence genes PWT3 and PWT4, whose gene products elicit defense in wheat cultivars containing the corresponding resistance genes Rwt3 and Rwt4. Studies on avirulence and resistance gene distributions, together with historical data on wheat cultivation in Brazil, suggest that wheat blast emerged due to widespread deployment of rwt3 wheat (susceptible to Lolium isolates), followed by the loss of function of PWT3. This implies that the rwt3 wheat served as a springboard for the host jump to common wheat.

Host jumps of plant pathogens may cause outbreaks of new crop diseases. A recent, likely example is wheat blast caused by Pyricularia oryzae (Magnaporthe oryzae). P. oryzae is composed of host-specific subgroups such as the Oryza, Eleusine, Avena, and Lolium pathotypes that cause disease in rice, finger millet, oat, and perennial ryegrass, respectively (13). Wheat blast was first reported in 1985 in Paraná state in the southern region of Brazil; it then spread to wheat-growing areas in neighboring states and countries in South America (4). The causal agent (the wheat blast pathogen) was identified as P. oryzae (4) but was specifically pathogenic on wheat and its wild relatives (Triticum spp. and Aegilops spp.) (3, 5); therefore, the pathogen is considered to be a previously unrecognized host-specific subgroup (Triticum pathotype). In 2011, wheat blast was found in North America, at a University of Kentucky research plot (6). Though less virulent than South American isolates, the strain isolated in Kentucky was pathogenic on wheat and was inferred to have evolved from annual ryegrass pathogen in North America through host jump (6). In 2016, wheat blast suddenly appeared in Bangladesh and caused a substantial loss of wheat production (7, 8). Phylogenomic analyses revealed that this outbreak was most likely caused by a wheat-infecting strain from South America (7, 8). Currently, this devastating disease has taken a major step toward becoming pandemic and poses a serious threat to global wheat production. Here we show that the wheat blast pathogen evolved through functional losses in a host specificity determinant.

Triticum isolates are most closely related to Avena and Lolium isolates (Fig. 1A). Takabayashi et al. (9) identified two genes, PWT3 and PWT4, in Avena isolate Br58, which conditioned its avirulence on wheat. They also identified the resistance gene in wheat that recognizes PWT4 and designated it as Rmg1 (synonymous with Rwt4). Similarly, Vy et al. (10) identified a gene (tentatively named A1) playing the primary role in the avirulence of Lolium isolate TP2 on wheat, as well as its corresponding resistance gene Rmg6. Allelism tests revealed that PWT3 is located at the same locus as A1 (fig. S1) and is recognized by Rmg6 (fig. S2). These data suggest that the same gene pair, PWT3 and Rmg6 (syn. Rwt3), is involved in the incompatibility of both Lolium and Avena isolates on wheat (fig. S3).

Fig. 1 Distribution of PWT3 and PWT4 in Pyricularia spp.

(A) Maximum likelihood tree of P. oryzae isolates constructed from SNPs in whole-genome sequences. P. grisea (Dig41) was used as an outgroup. The numbers on the branches indicate bootstrap probability. The bar below the tree indicates genetic distance per site. (B) Schematic representation of PWT3 and PWT4 types among Triticum isolates (highlighted in the shaded region) from the 1990s (Kobe University collection) and other pathotypes. The arrows and horizontal bars represent ORFs and flanking regions, respectively. For PWT3, Ao and B are avirulent and virulent types, respectively. The A′ and Atm types are identical to the Ao type, except for a one-base substitution and an insertion of reprotransposon MGR583 (gray triangle) in the upstream region, respectively. For PWT4, red, green, and blue arrows represent the avirulent type, the virulent type, and a truncated virulent type, respectively. Dotted lines indicate the absence of homologs. bp, base pairs.

We isolated PWT3 from Avena isolate Br58 through map-based cloning (fig. S4). The PWT3 nucleotide sequence from Br58 (later designated as A type) was shared by all 12 Lolium isolates analyzed, including TP2 (table S1), supporting the status of PWT3 as a host species specificity gene. We also isolated PWT4 from Br58 by using bulked segregant analysis coupled with whole-genome sequencing (fig. S4). The predicted PWT3 and PWT4 proteins contained putative signal peptides (fig. S4) but lacked similarity to known proteins or protein domains.

PWT3 and PWT4 were identified through seedling infection assays, but wheat blast is mainly a spike disease in Brazilian wheat fields. To estimate the roles of these genes during field infections, spikes of wheat cultivars Norin 4 (N4) (Rwt3/Rwt4), Chinese Spring (CS) (Rwt3/rwt4), Transfed (Tfed) (rwt3/Rwt4), and Hope (rwt3/rwt4) were inoculated at early anthesis with Triticum isolate Br48 and transformants carrying PWT3 (Br48+3) and PWT4 (Br48+4). Br48 was virulent on spikes of all four cultivars, whereas Br48+3 and Br48+4 showed specific avirulence on the Rwt3 and Rwt4 carriers, respectively (Fig. 2A). Thus, PWT3-Rwt3 and PWT4-Rwt4 interactions play critical roles in spike infection.

Fig. 2 PWT3 and PWT4 serve as the host species specificity barrier for wheat.

Spikes of wheat cultivars Norin 4 (N4), Chinese Spring (CS), Transfed (Tfed), and Hope were inoculated with wild types (WT) of Triticum (A), Lolium (B), and Avena (C) isolates; their transformants carrying introduced PWT3 (+3) or PWT4 (+4); and disruptants of PWT3 (Δ3), PWT4 (Δ4), or double disruptant (Δ3Δ4). Inoculated spikes were incubated for 8 days.

We performed gene-disruption experiments to determine whether PWT3-Rwt3 and PWT4-Rwt4 interactions are the only barriers preventing the nonwheat isolates from infecting Rwt3/Rwt4 wheat. Lolium isolate TP2 was avirulent on Rwt3 cultivars but virulent on rwt3 cultivars (Fig. 2B). When PWT3 was disrupted (fig. S5), the resulting strain TP2Δ3 gained virulence on the Rwt3 cultivars (Fig. 2B and fig. S6). Avena isolate Br58 was avirulent on the Rwt3 and/or Rwt4 carriers and virulent only on Hope (rwt3/rwt4) (Fig. 2C). When PWT3 and PWT4 were disrupted individually (figs. S5 and S7), the resulting strains Br58Δ3 and Br58Δ4 gained virulence on CS (Rwt3/rwt4) and Tfed (rwt3/Rwt4), respectively (Fig. 2C and fig. S6). Furthermore, a double disruptant gained virulence on all cultivars, including N4 (Rwt3/Rwt4). These results indicate that mutations or deletions of PWT3 and PWT4 would lead to a gain of virulence on a majority of wheat cultivars.

To determine whether mutation or loss of PWT3 and/or PWT4 served as key events in the evolution of wheat blast, we screened for their presence in genome sequences from a comprehensive collection of P. oryzae strains. PWT4 homologs were found in some isolates but were considered nonfunctional (Fig. 1B and fig. S8). In contrast, PWT3 homologs were detected in all isolates tested, with the open reading frame (ORF) sequences conforming to one of four basic types (A, B, C, and D) (Fig. 1B and fig. S9). The Avena and Lolium A-type homolog was also found in Eleusine isolates and some Triticum isolates (Fig. 1B). Upstream sequences further divided the A type into three subtypes (Ao, Ao′, and Atm). Compared with the original Br58 type (Ao), the Eleusine isolates’ type (Ao′) had one base substitution in the upstream region. Three Triticum isolates (Fig. 1) represented three distinct types: the Ao type; the Atm type, carrying an insertion of the LINE-like element MGR583 (11) (fig. S10A); and the B type, which had 13 single-nucleotide polymorphisms (SNPs) causing 12 amino acid substitutions in ORFs (fig. S9A). Among 19 Triticum isolates collected in Brazil from 1990 to 1992, six, six, and seven isolates contained the Ao, Atm, and B types, respectively (table S2).

The Atm and B carriers were virulent on the four test cultivars, whereas the Ao carriers were avirulent on the Rwt3 cultivars (table S2 and fig. S6A). An in planta expression analysis revealed that the Ao- and B-type genes in Triticum isolates were as highly expressed as the Ao type in Br58 (fig. S10B). In contrast, the expression level of the Atm type was reduced by a factor of 10 (fig. S10B). These results indicate that the MGR583 insertion in the Atm type has compromised its avirulence function by reducing expression.

To determine the origins of the highly divergent PWT3 types, we screened ~100 representative Pyricularia isolates from various hosts and found that an isolate from Brachiaria plantaginea (Br35) collected in Brazil in 1990 carried the B-type PWT3 with 100% nucleotide sequence identity to that of Br48. However, it seemed unlikely that such a Brachiaria isolate was a direct ancestor of Br48 because Br35 was phylogenetically remote from Br48 (fig. S11). Comparative analyses of whole-genome sequences suggested that the direct ancestor of Br48 inherited a 1.6-Mb chromosomal segment carrying the B-type PWT3 from a Br35-like Brachiaria isolate (fig. S12).

The Triticum isolates employed above were early isolates collected from 1990 to 1992 (Kobe University collection). To reveal population dynamics, we performed PWT3 typing with more recent Triticum isolates preserved at universities or institutes in the Americas and genome sequences in public databases. Kentucky strain WBKY11-15 collected from wheat in 2011 had a previously unrecognized type (Atp) composed of an A-type PWT3 ORF and an inverted repeat transposon Pot3 (12) inserted into its upstream region (Fig. 3A). Bolivian strain B71 collected in 2012 had another type (Atc) in which the PWT3-ORF (with two nucleotide substitutions, in comparison with the A type) was disrupted by a complex insertion of transposable elements, Pyret (13) and RETRO5 (14) (Fig. 3B). All isolates characterized from Bangladesh in 2016 carried the Atc type (Fig. 3C). The Atc type was present in the 1990s as a minor population in Brazil’s southern states (Fig. 3C) and became prevalent throughout Brazil in the 2000s. It spread to other countries in South America in the 2010s and was finally transmitted to Asia and caused the outbreak of wheat blast in Bangladesh.

Fig. 3 Dynamics of the PWT3 types during three decades.

(A) Structure of the Atp type in WBKY11-15 (LC229726), an isolate collected from wheat in Kentucky. (B) Structure of the Atc type in B71 (LC215054), a highly aggressive Triticum isolate collected in Bolivia. D, Asp; N, Asn; V, Val; I, Ile. (C) Geographical and historical distribution of the PWT3 types in Triticum isolates. Each symbol represents an isolate collected in the indicated year and is color-coded according to its PWT3 type. The Atc type appears to have expanded over time, whereas the B type was not found beyond the 1990s.

A critical issue is why strains with the functional Ao type PWT3 have been isolated as Triticum isolates. To answer this question, we surveyed the distribution of Rwt3 and Rwt4 in common wheat (fig. S13). In 499 local landraces collected worldwide, Rwt3 and Rwt4 carriers accounted for 77 and 87%, respectively (Fig. 4 and table S3). Only 6.6% of the accessions lacked both genes. Such ubiquity appears to be a key characteristic that distinguishes resistance genes conditioning the subgroup-genus specificity from those conditioning race-cultivar specificity. When ~60 improved cultivars from the Americas were tested, however, we found a noteworthy change in the early 1980s. In the late 1970s and early 1980s, the most planted cultivar in Brazil was IAC-5 (15), a carrier of Rwt3 (table S4). Around 1980, a new cultivar (Anahuac) was introduced to Brazil and recommended to farmers (15) because it adapted very well in nonacidic soils. Anahuac was a semidwarf cultivar with high yield potential but a noncarrier of Rwt3 (table S4). In 1985, a few years after the release of Anahuac, the outbreak of wheat blast occurred (4). Such wide cultivation of rwt3 cultivar(s) in Brazil would explain why P. oryzae strains with the Ao type PWT3 have been isolated from common wheat in the early period of the wheat blast outbreak in Brazil (table S2) and thereafter (Fig. 3C).

Fig. 4 Global distribution of Rwt3 and Rwt4 in local landraces of common wheat.

Pie chart size is in proportion to numbers of accessions tested. The inset is a magnified map of the region inside the dashed box.

On the basis of the results described above, we present a model for the emergence of the wheat blast fungus in Brazil (fig. S14). Widespread cultivation of rwt3 cultivars in the early 1980s allowed certain P. oryzae strains to colonize common wheat and increase the population despite carrying the Ao type PWT3. Wheat cultivars possessing Rwt3 were still cultivated nearby and imposed selection for mutation or loss of PWT3. As a result, pwt3 strains arose through independent events involving either de novo transposon insertion or gain of mutated PWT3 from a remote strain, which would finally establish wheat pathogens as pathogenic to the entire wheat population.

This model implies that rwt3 cultivars served as springboards for the host jump of P. oryzae strains (with PWT3;pwt4) in Brazil. Similarly, rwt3/rwt4 cultivars may become springboards for Avena isolates (PWT3;PWT4) to jump hosts. Taken together, it is advisable to cultivate common wheat cultivars that carry both Rwt3 and Rwt4 for forestalling reoccurrence of host jumps or at least for preventing wheat blast disease caused by infection with Lolium or Avena pathogens.

Supplementary Materials

Materials and Methods

Figs. S1 to S14

Tables S1 to S4

References (1631)

References and Notes

Acknowledgments: We thank S. Kamoun (The Sainsbury Laboratory), T. Wolpert (Oregon State University), A. S. Urashima (Federal University of Sao Carlos), Y. Takano (Kyoto University), and H. Nakayashiki and K. Ikeda (Kobe University) for comments on drafts of the manuscript and S. Liu (Kansas State University) for sharing PacBio sequence data for the complex transposon insertion. Nucleotide sequence data reported herein are available in the DNA Data Bank of Japan (DDBJ) Sequenced Read Archive under accession number DRA005349 and in the DDBJ, European Molecular Biology Laboratory, and GenBank databases under accession numbers LC202650 to LC202657, LC215053, LC215054, and LC229726. This project was supported by Japan Society for the Promotion of Science grant 26292025, the Agriculture and Food Research Initiative competitive grant 2013-68004-20378 from the U.S. Department of Agriculture National Institute of Food and Agriculture, and the Hatch project KY012037 under accession number 1002523. This is contribution number 17-356-J from the Kansas Agricultural Experiment Station and publication number 17-12-051 of the Kentucky Agricultural Experiment Station. The supplementary materials contain additional data.

Stay Connected to Science

Navigate This Article