We have determined the genomic sequence of the wheat FHB resistance QTL Fhb1 in the resistant donor cultivar CM-82036 and positioned the gene within a region covering 860 kb using a fine-mapping panel for Fhb1 that has been phenotyped for both FHB and DON resistance. 28 candidate genes including 13 high confidence genes located in this interval have been further characterized in a dense time-course RNA-seq study.
Suppressed recombination in the highly divergent region of Fhb1
Our findings show that the sequence containing Fhb1 in CM-82036 differs significantly from the susceptible Chinese Spring reference in gene content and size. The core region of the sequenced contig is highly dissimilar to Chinese Spring when comparing genomic DNA; also markers designed for this region in CM-82036 failed to amplify in the susceptible region of cv. Remus. A dot plot analysis did not identify structural rearrangements such as genomic inversions or duplications (Fig. 1b). The in part large differences in distance between the genes present in resistant and susceptible cultivars may be attributed to different transposon insertions events (Scherrer et al. 2005). Most genes unique for Fhb1 reside in a compact cluster (genes #18–#27). The observed differences can be explained in part by pseudogenisation of genes #21, #26 and #27 in Chinese Spring where partial overlaps were still identifiable in non-annotated regions. However, multiple genes are unique for either the Fhb1 contig or Chinese Spring.
Haplotype divergences and loss of microcolinearity between cultivars on a comparable scale as detected for Fhb1 have been observed several times before (i.e., for the barley Rph7 locus (Scherrer et al. 2005) and wheat Lr10 (Isidore et al. 2005)) and also in two recent studies: Yeo et al. (2015) have resequenced the resistant and susceptible loci from two barley cultivars differing in the Puccinia hordei resistance gene Rphq2 and identified entirely different haplotypes of which the resistant locus harbors unique candidate genes. Mago et al. (2014) established the genomic region harboring the wheat stem rust resistance gene Sr2, which includes a cluster of germin-like proteins missing in the susceptible reference cultivar Chinese Spring. The authors showed that this cluster is shared between other resistant accessions containing Sr2.
In rice, more than 6 % of the genome is occupied by regions of high divergence (Tang et al. 2006). The higher genome plasticity of polyploid wheat led to a high rate of gene deletions and activity of repetitive elements buffered by the redundancies within the three homoeoallelic subgenomes (Dubcovsky and Dvorak 2007). Most likely the rate of highly divergent regions between cultivars is more frequent than generally assumed. Redundant gene content may also more easily accommodate introgressions of rare resistance haplotypes from landraces under selective pressure by replacing genes with redundant gene activity.
Highly diverging haplotypes generally lead to strongly reduced meiotic recombinations in such regions. In the Fhb1 locus, loss of colinearity has a direct effect on recombination frequency. While flanking recombining regions harbor 4 and 6 recombinations in an interval of 42.6 and 122 kb, respectively (Fig. 3) about 700 kb remain unresolved. On a larger scale, the locus itself resides in a highly recombining telomeric region of chromosome 3B with an average of 0.85 cM/Mb (Saintenac et al. 2009), which would relate to 17 expected recombinations within the Fhb1 non-recombining region. Tracking recombinants from two mapping populations of which neither parent harbored Fhb1 the same authors detected a recombination hotspot that covers the region around UMN10 in one of their populations leading to high recombination rates, while only a below average rate was observed in the second population (Saintenac et al. 2011). Apparently, cross-over hotspots exist in this region, but these need to be met with matching crossing partners.
As CM-82036 is a direct derivative from a cross of the Fhb1 donor Sumai-3, the sequence obtained from this cultivar should be directly comparable to the sequence of Sumai-3, which has been target of several fine-mapping studies (Cuthbert et al. 2006; Liu et al. 2008; Bernardo et al. 2011). Among the markers used in both studies, the widely used UMN10 is the only marker we found mapping within the non-colinear region, while all others mapped to flanking regions. The original UMN10 from Liu et al. (2008) detects a length polymorphism. This marker has been converted into an easily applicable KASP-SNP assay, which has been successfully employed on diverse germplasm conducted by several groups so far and should remain a reliable marker for broader use.
The closest reported recombination event is based on a single recombinant line for marker sts32 reported by Liu et al. (2008) which mapped to position 155.4 kb on the Fhb1 contig. Consideration of this single event would exclude genes #1–#10 from the list of Fhb1 candidates. To further substantiate these findings, we identified 100 recombinant lines in our own fine-mapping panel for the gwm493 and barc133 interval. Yet, we failed to identify additional recombinant lines that would reach as far or further in the ‘core’ region of Fhb1.
Candidate genes in the Fhb1 locus
Fhb1 expresses a dominant phenotype (Xie et al. 2007 and own unpublished data). Consequently, possible explanations for the phenotypic difference could be induced expression of the underlying gene in response to the pathogen or constitutive expression in the resistant genotype, or absence of the respective gene in susceptible lines as Chinese Spring and Remus. Furthermore, gain of function polymorphisms through changes in protein sequence may cause the resistance phenotype. Also susceptibility factors encoded in the interval in lines lacking Fhb1 need to be considered, although such a scenario is more difficult to reconcile with the reported dominance. Our results also show that DON resistance (determined as bleaching resistance after application of high concentrations of pure toxin) is clearly associated with Fhb1. Either DON resistance itself could simultaneously lead to FHB resistance, or the gene causing DON resistance might be tightly linked to the gene conferring FHB resistance by a different mechanism. The fact that increased F. graminearum resistance was achieved by increasing DON resistance [due to overexpression of a barley glucosyltransferase (Li et al. 2015)] suggests that genes with an effect on toxin resistance should be considered as prime candidate Fhb1 genes. With the sequenced region at hand and mapped expression data, these scenarios can now be considered much better; albeit lacking a higher resolved map still many candidate genes remain: We discuss putative functions and expression patterns of the candidate genes and the implications for functional testing.
Gene #17 (ubiquitin-2 like Rad60 SUMO-like protein) is unique for the CM-82036 sequence, deleted in the susceptible cultivar (see Fig. 1c). In yeast, it has been shown that reducing the ubiquitin pool by disruption of the stress responsive polyubiquitin gene leads to reduced DON resistance of ubi4 mutants (Abolmaali et al. 2008). Yet, gene #17 is practically not expressed, neither in the control nor following F. graminearum-inoculation. Similarly, genes #21 (general transcription factor IIE subunit) and #25 (cystatin) are not present in the susceptible Chinese Spring reference, but also not expressed under both conditions. We, therefore, consider them unlikely candidates for Fhb1.
All other genes within the diverging region are expressed in the Fhb1-containing CM-NIL38 but not in the susceptible CM-NIL51. Expression levels range from few reads per sequenced sample to hundreds of reads per sample. While highly expressed genes may present themselves as ‘more-likely’ candidates, comparably lowly expressed genes should not be ruled out as RNA levels may not directly reflect protein expression levels and the abundances to establish a specific function may be vastly different for individual gene products.
Gene #19 (terpene synthase), gene #20 (unknown protein) and gene #23 (E3-Ubiquitin ligase) are also present in Chinese Spring but they are only expressed in Fhb1 containing lines. Terpene synthases act in the biosynthesis of secondary metabolites, which play a role in defense against herbivores or pathogens (Lange 2015). Many terpenoid phytoalexins from Poaceae have been described (Ejike et al. 2013). These include phytoalexins derived from monoterpenes and sesquiterpenes, which have a direct antimicrobial effect (Schmelz et al. 2011; Inoue et al. 2013). Sesquiterpenoid phytoalexins active against F. graminearum have for instance been described in maize (Huffaker et al. 2011). Volatile terpenes also may act as messengers upon pathogen attack (Nagegowda 2010). The Fhb1-associated terpene synthase most likely acts in synthesis of cytosolic sesquiterpenoids from farnesyl diphosphate as a BLASTp result suggests (delta-cadinene synthase isozyme A, e value = 0). However, the overall low expression level of the terpene synthase encoded on Fhb1, with no observable differences between F. graminearum treatment and mock, suggests that this gene does not play an active role in the response to the pathogen.
A secondary annotation for gene #20 (unknown protein) suggests a role in calcium sensing (sarcoplasmic reticulum histidine-rich calcium-binding protein precursor, blastp, e value = 8e−38) and, consequently, may lead to changes in gene expression following external cues such as abiotic and biotic stresses (Reddy et al. 2011). Yet, its strong constitutive expression suggests a different role for this gene. The lack of expression in the susceptible NIL is not due to pseudogenisation of the CS ortholog. Both gene models seem intact and share high amino acid sequence similarity (95.7 %, Supplementary File Fig. S6). In contrast, the promotor regions are highly divergent in the first 1 kb upstream of the short 5’UTR region with multiple sequence deletions in the Fhb1-contig, which most likely are responsible for the observed large expression differences. Despite the lack of clear indications about its potential mode of action, gene #20 should be considered a candidate for Fhb1.
Gene #23 is a predicted E3-ubiquitin ligase of the ‘seven in absentia’ (SINA) type. Such proteins mediate ubiquitination and proteasome-mediated degradation of specific proteins (in response to a stimulus). Some SINA proteins and their client proteins have been implicated in plant–pathogen and plant–symbiont interactions (Kim et al. 2006; Den Herder et al. 2012). They contain an N-terminal RING domain and a C-terminal conserved domain implicated in dimerization and substrate binding. Interestingly, the gene models for this protein differ largely due to an internal deletion of 31 nt in the Fhb1 reading frame compared to the gene model in Chinese Spring. The consequence is a frameshift and premature stop codon, removing the entire SINA domain. The Fhb1-resistant line, therefore, possesses most likely a nonfunctional version of the protein. Yet, the expression is higher in the Fhb1 background, so potentially the truncated protein might act in a dominant negative fashion, so that it may be premature to exclude this gene as a candidate.
The Fhb1 region hosts four clearly expressed genes absent in the susceptible reference
Gene #22 encodes protein with domains encoding agglutinin and ‘pore-forming toxin-like’. This weakly expressed gene might have direct antifungal activity by binding to fungal cell wall carbohydrate structures and permeating membranes. The role of lectins in plant defense is well established (Lannoo and Van Damme 2014). Wheat germ agglutinin has been shown to bind to N-acetyl-d-glucosamine (Levy 1979), a monomer of the fungal cell wall chitin and as such constitutes a pathogen recognition mechanism that elicits further, early defense responses. Wheat germ agglutinin exhibits also a negative effect on hyphal growth of various fungi including F. graminearum (Mirelman et al. 1975; Ciopraga et al. 1999). This proposed mechanism is, however, more consistent with type I than type II resistance against spreading of the disease and resistance to DON. Expression of the gene #22 cDNA in Saccharomyces cerevisiae under control of the inducible GAL1 promoter did not affect the growth of the transformed yeast strain on galactose medium (data not shown).
Gene #24 encoding a GDSL lipase is the only gene in the sequenced contig that exhibits a significant increase in expression in response to the pathogen. GDSL lipase/esterases comprise a structurally diverse gene family in plants. For instance, 114 members exist in the rice genome (Chepyshko et al. 2012). They act in regulation of a variety of physiological functions including defense. A chain of studies (Kwon et al. 2009; Kim et al. 2014) demonstrated the role of an Arabidopsis thaliana GDSL lipase 1 in modulating systemic immunity through the regulation of ethylene signaling in response to necrotrophic pathogens. However, the present GDSL lipase does not share similarity with the A. thaliana lipase 1 gene. The expression pattern and its possible role in defense warrant further investigations.
Gene #26 (F-box protein) is among the strongest constitutively expressed genes on the contig. No similar gene is annotated in this region of Chinese Spring, yet mapping of the coding sequence of this gene onto Chinese Spring identified a likely pseudogene with weak similarity (Table 1). F-box proteins are part of the ubiquitination complex, which form specific interaction with target proteins. Consequently, the gene family is very large with 779 genes in rice (Xu et al. 2009). The F-box protein could be involved in reducing the levels of protein encoding a susceptible factor for FHB. Its target protein, which is most likely not encoded in the Fhb1 region, would need to be genetically fixed and must not segregate, to be in agreement with the absence of epistasis at Fhb1. Potentially, the F-box protein could also directly target an unknown effector protein of the pathogen. In A. thaliana, the F-box protein encoded by COI1 is involved in jasmonate signaling and is the target of the jasmonic acid mimicking bacterial toxin coronatine which increases susceptibility (Geng et al. 2012).
Also gene #27 (hypothetical protein) cannot be excluded as gene candidate. However, only few reads map to the predicted CDS of this low confidence gene for which no annotation could be retrieved.
The genes on the right half of the Fhb1 interval (Fig. 1c) have again counterparts in the susceptible line. The genes #28 and #29 are constitutively expressed and are discussed below. Gene #30 encodes a predicted zinc finger C3H4 type (RING finger) domain-containing protein showing low expression in both NILs, and no response to F. graminearum infection. Zinc finger-containing proteins have functions ranging from transcription, translation, mRNA trafficking, cytoskeleton organization, protein folding, chromatin remodeling and more. Only a domain of unknown function (DUF3675) is additionally recognized. But since the gene model is identical with that of Chinese Spring, this gene showing no significant expression difference between NILs and in response to F. graminearum can be excluded.
Also genes #31 and #32 are unlikely candidates due to lacking expression. A CYP450 gene could encode an enzyme involved in the biosynthesis of an antifungal metabolite, or a detoxification enzyme leading to chemical modification of the toxin structure. A bacterial cytochrome P450 detoxifying DON by hydroxylation of C16 has been described (Ito et al. 2013). The product of gene #32 contains an NB-ARC domain, which is found in plant disease resistance genes (van der Biezen and Jones 1998). Besides the nucleotide binding domain, also leucine-rich repeats can be recognized. A highly similar protein from Aegilops tauschii has been annotated as ‘putative disease resistance RPP13-like protein 1’ (GenBank accession: EMT27135.1). The version of the susceptible Chinese spring gene is identical in 898 of 905 amino-acids, leaving room for functional differences (Supplementary File Fig. S7). Yet, lack of expression is hard to reconcile with the otherwise suggestive role of this candidate disease resistance gene.
How can genes on the Fhb1 contig help explain the higher ability to inactivate DON?
Lemmens et al. (2005) have associated the Fhb1 locus with the higher ability to metabolize DON into the non-toxic DON-3-O-glycoside, which is a product of the activity of toxin-specific UDP-glucosyltransferases (UGT, Poppenberger et al. 2003). No such gene is encoded on the Fhb1 contig, gene #6 annotated as a HGA-like UGT does share similarities to the large super family encoding small molecule conjugating UGTs (Ross et al. 2001), but most likely acts on the formation of homogalacturonan (HGA) as part of the cell wall (Yin et al. 2010). Toxicity of DON is caused by inhibition of protein biosynthesis; therefore, genes involved in translation may counteract the adverse effect of DON by increasing overall translation fidelity or exerting a greater tolerance to DON in other ways. Genes #7 and #9 encode leucyl- and alanyl-tRNA synthases, respectively. While gene #7 shows no significant expression difference, gene #9 is clearly higher expressed in the Fhb1-containing NIL (Fig. 2) and is, therefore, more attractive. In addition, potentially relevant sequence differences exist (Supplementary File Fig. S8). Recently, it has been shown that overexpression of a methionyl-tRNA synthase from wheat when overexpressed in A. thaliana causes increased DON resistance in transformants (Zuo et al. 2016). Yet, as stated above, if the reported single recombinant line at sts32 (Liu et al. 2008) is indeed correct, all genes up to #10 can be ruled out as candidates. Gene #13 (tRNA-modifying methyltransferase) has a ribosome-associated function, likewise #28 (translation initiation factor). A possible role of this gene for methylation-associated resistance of ribosomes against trichothecene toxins has been proposed by Iglesias and Ballesta (1994), who found that in Fusarium oxysporum adaptive toxin resistance of ribosomes can be obtained by enzymatic modification of an unknown ribosomal component upon incubation with S-adenosylmethionin. Despite its low expression in the Fhb1-containing NIL, this gene should not be ruled out as candidate. Also #29 has a predicted methyltransferase domain.
With the sequence of Fhb1 at hand the genes described in this study are a valuable resource for further functional analysis of the QTL. Based on expression profiles and annotations, some genes can be ruled out, but many remain for which further functional assessments are required. The knowledge of which gene is causing FHB resistance is not irrelevant, as breeders unknowingly may deploy proteins with potentially undesired health effects (lectin/pore-forming toxin) or increase the levels of antifungal compounds with unknown toxicological properties (terpenoid synthase). The most promising approach to further characterize Fhb1 is the characterization of EMS-generated stable loss-of-function mutants (Slade and Knauf 2005) for which polyploid wheat is especially well suited due to the high possible mutation rates. RNA-interference methods such as VIGS may not yield clear phenotypes for targeted candidate genes as the silencing is only partial and transient. This residual expression levels bear the risk of providing sufficiently high mRNA levels to produce relevant amounts of protein to confer the resistance phenotype. Generating stable wheat transformants in a type 2 susceptible cultivar is a viable alternative to assess candidates for FHB resistance (Li et al. 2015) and should bring clarity about the gene underlying FHB and DON resistance. While the simplest hypothesis is that only one gene is causing both phenotypes, also the scenario of two different resistance genes cosegregating due to repressed recombination in the region cannot be excluded.
Author contribution statement
Generation of plant material: GS, BS. BAC library construction, screening and sequencing: SV, WS. Contig assembly and annotation: WS. Greenhouse trials: BS, FJ, MZ. Toxin and strain provision: ML. Marker design and genotyping: BS, VG, FJ, WS. RNAseq data acquisition and analysis: MZ, WS, TM. HB (BOKU), KFXM, GA and HB (INRA) conceived this study and obtained funding. Manuscript writing: WS, BS and GA. The manuscript was finally approved by all coauthors.