Resistance to wilt fungus Fusarium oxysporum f.sp. matthioli (FOM) is a polygenic trait in Arabidopsis thaliana. RFO3 is one of six quantitative trait loci accounting for the complete resistance of accession Columbia-0 (Col-0) and susceptibility of accession Taynuilt-0 (Ty-0).
We find that Col-0 and Ty-0 alleles of RFO3 are representative of two common variants in wild Arabidopsis accessions, that resistance and susceptibility to FOM are ancestral features of the two variants and that resistance from RFO3 is unrivalled by other genes in a genome-wide survey of diversity in accessions. A single receptor-like kinase (RLK) gene in Col-0 is responsible for the resistance of RFO3, although the susceptible Ty-0 allele codes for two RLK homologs.
Expression of RFO3 is highest in vascular tissue, which F. oxysporum infects, and root-expressed RFO3 restricts FOM infection of the vascular system. RFO3 confers specific resistance to FOM and provides no resistance to two other crucifer-infecting F. oxysporum pathogens.
RFO3's identity, expression and specificity suggest that RFO3 represents diversity in pattern-recognition receptor (PRR) genes. The characteristics of RFO3 and the previously published RFO1 suggest that diversity in RLK PRRs is a major determinant of quantitative resistance in wild plant populations.
Innate resistance to infectious disease varies among individuals of the same species (Laine et al., 2011). Plant breeders exploit the genes controlling this natural variation in resistance, so-called resistance genes, in available germplasm to improve host resistance of crops (Boyd et al., 2012). In crosses between resistant and susceptible varieties, qualitative resistance may segregate as a monogenic trait, or as a combination of multiple quantitative trait loci (QTLs) conferring polygenic resistance (St Clair, 2010). Most research and breeding programs favor working with the discontinuous resistance and simple Mendelian inheritance of monogenic traits. However, outside of laboratories and in natural populations, QTLs may well account for much of the variation in disease resistance, as polygenic resistance is prevalent and often complicates the analysis of Mendelian ratios (Kou & Wang, 2012).
Most identified resistance genes confer strong resistance and the hypersensitive response to specific pathogenic strains and typically encode members of the nucleotide binding, leucine-rich repeat (NB-LRR) class of resistance proteins (McHale et al., 2009). On the other hand, few reports have identified resistance genes conferring quantitative resistance, and whether particular gene families are more often associated with quantitative traits remains speculative (Kou & Wang, 2012; Poland et al., 2009; St Clair, 2010). Also, with little molecular description of quantitative resistance, it remains unclear whether and how quantitative resistance is divided into distinct types.
Fusarium wilt of Arabidopsis thaliana is an experimental pathosystem for studying the genetic nature of polygenic resistance and interaction of a plant host with vascular wilt fungi (Diener & Ausubel, 2005; Diener, 2012). Fusarium wilt diseases limit cultivation of numerous crops that are susceptible to vascular infection by pathogenic forms, or formae speciales, of common soil fungus Fusarium oxysporum (Mace et al., 1981; Kistler, 1997). A. thaliana is susceptible to three crucifer-infecting formae speciales, namely F. oxysporum f.sp. matthioli (FOM) isolated from garden stock (Matthiola incana), F. oxysporum f.sp. conglutinans (FOC) isolated from diseased Brassica spp. and F. oxysporum f.sp. raphani (FOR) isolated from diseased radish (Raphanus sativus) (Bosland & Williams, 1987; Diener & Ausubel, 2005). Infected A. thaliana recapitulates the foliar wilt symptoms observed in native field hosts, such as stunting, epinasty, yellowing and premature senescence of leaves.
Three phases describe interactions between vascular wilt fungi and their hosts (Talboys, 1972; Diener, 2012). In the primary determinative phase, F. oxysporum penetrates and colonizes extravascular tissue usually at growing root tips. In the secondary determinative phase, F. oxysporum enters the vascular cylinder and spreads through water-conducting xylem vessels. In the expressive phase, symptoms become evident in the above-ground foliage and at a distance from the site of infection. The rapidity and strength of the host's response to infection in the first two phases determine the effectiveness of resistance.
Resistance to FOM in A. thaliana is a polygenic trait (Diener & Ausubel, 2005). The response of wild accessions of A. thaliana to infection by FOC, FOM or FOR ranges widely from complete resistance to ready susceptibility. For instance, accession Col-0 exhibits complete resistance to a dose of FOM that consistently kills accession Taynuilt-0 (Ty-0). On the other hand, Ty-0 and Col-0 exhibit similar partial resistance when accessions are instead infected with FOC race 1. Thus, in large part, resistance is quantitative and specific to the infecting forma specialis. In the cross between Col-0 and Ty-0, six RESISTANCE TO F. OXYSPORUM (RFO) QTLs account for complete resistance of Col-0 and relative susceptibility of Ty-0. One QTL, RFO1, encodes a member of the wall-associated kinase family of receptor-like kinases (RLKs) and quantitatively contributes to immunity as loss-of-function in rfo1 enhances F. oxysporum infection in the root vascular cylinder (Diener & Ausubel, 2005; Diener, 2012).
Here we attribute quantitative variation in resistance to diversity in members of the S domain subfamily 1 (SD1) of RLK genes at the RFO3 locus. RFO3's identity, effect, and specificity imply that RFO3 functions as a PRR. The coincidence that two RFO QTLs, RFO3 and the previously described RFO1, correspond to diversity in RLK genes argues that putative PRR RLK genes are a major source of quantitative variation in disease resistance in wild plant populations.
Materials and Methods
Taynuilt-0 (CS6768), other Arabidopsis accessions (CS22660), Sail_1212_G06 (CS844287) and BAC clones MVC8 and MSL1 were purchased from the Arabidopsis Biological Resource Center (Ohio State University, Columbus, OH, USA). rfo1 is from Diener & Ausubel (2005); and, 6E5 (RFO3-C/T) is a F1BC plant from the (Col-0 × Ty-0) × Ty-0 mapping population in Diener & Ausubel (2005). Arabidopsis was grown on Jiffy7 peat pellets, 1″ × 1.25″ #734 or 1.5″ × 1.5″ #730 (Growers Solution, Cookeville, TN, USA) (#730), under moderate cool white fluorescent lighting with 12-h day length at 28°C and night temperature of 25°C. For axenic growth of seedlings, bleach-sterilized seeds were sown on Petri plates with plant nutrient minimal medium and 0.8% agar alone or, for antibiotic selection, with 0.5% sucrose (Haughn & Somerville, 1986). Transgenic plants were selected on kanamycin (20 mg l−1). Petri plates were sealed with Micropore tape (3M Corp., St Paul, MN, USA). Single-hypocotyl grafts of young seedlings were performed as described in Turnbull et al. (2002).
F. oxysporum infection
Strains of FOC race 1 (777), FOC race 2 (808), FOR (815) and FOM (726) originate from P.H. Williams through H.C. Kistler (Bosland & Williams, 1987; Kistler et al., 1991). K. O'Donnell (ARS, USDA, Peoria, IL, USA) provided FOM isolates NRRL-22545, -38343, -38339 and -38360 (O'Donnell et al., 2009). F. oxysporum cultures were grown and harvested as in Diener & Ausubel (2005). For infection, conidial density was adjusted to between 106 and 108 conidia ml−1 using a hemacytometer, and 2 to 3-wk old plants were irrigated with conidial suspension or water (for controls). Infected plants were scored using a health index (HI) with ratings of zero (dead) to five (unaffected), which is the same as disease index in Diener & Ausubel (2005). For further discrimination of symptoms, ratings between whole numbers were recorded in 0.5 increments. If single HI scores of plants are reported, plants were susceptible if HI < 3, resistant if HI ≥ 4, or had an intermediate resistance if 3 ≤ HI < 4, respectively. Alternatively, infected plants within an experiment were rank-ordered from most susceptible to most resistant. Rank order was also derived from multiple HI scores from two or more time points, in which case later HI scores had priority over earlier HI scores. From rank order, plants with the lowest third of ranks, highest third of ranks or middle third of ranks were arbitrarily deemed susceptible, resistant or having intermediate resistance, respectively. Differences in median ranks of genotypes were evaluated using the Mann–Whitney rank sum test.
In F1BC plants of the (Col-0 × Ty-0) x Ty-0 mapping population in Diener & Ausubel (2005), recombination breakpoints between simple sequence length polymorphism (SSLP) markers CIW11 and nga162 on chromosome 3 located RFO3 between these markers. To assign RFO3 genotypes to plants with informative recombination breakpoints, the cosegregation of Rfo phenotype (quantitative resistance to FOM) and RFO3-linked marker genotype was tested in 25–50 F2BC progeny. If Rfo phenotype and marker genotype cosegregated, the genotype was RFO3-C/T; and, if Rfo phenotype and marker genotype assorted independently, the genotype was RFO3-T/T or RFO3-C/C. The few hundred F2BC surviving FOM infection were recycled to identify new recombinant breakpoints. Recombinant breakpoints in F2BC plants 1E2.A6 and 2C3.B9 put RFO3 between SSLP markers MJK13 and MSL1.60. The description of primers used for mapping are in the Supporting Information, Table S1. RFO3 genotypes of lines 1E2.A6 and 2C3.B9 were tested in the progeny of a backcross to Ty-0. Rfo phenotype and MVC8 marker genotype cosegregated in backcross progeny of both F2BC plants (as shown in Fig. S1A and C, respectively), and thus 1E2.A6 and 2C3.B9 were RFO3-C/T. In addition, we tested the subsequent generation from selfed heterozygous progeny of 1E2.A6 and 2C3.B9, and again RFO3-linked markers cosegregated with Rfo phenotype (as shown in Fig. S1B and D, respectively). A screen of 400 F2 plants of cross rfo1 (Col-0) × Ty-0 identified three plants, including 1D3, with recombination breakpoints between MVC8 and MSL1.29. 1D3 was RFO3-C/T as Rfo phenotype and RFO3-linked marker genotypes cosegregated in F3 of 1D3 (Fig. S1E).
Transformation was performed using the floral dip method and Agrobacterium tumefaciens strain GV3101 (Clough & Bent, 1998). T1 transformants were selected for kanamycin resistance. Pure-breeding kanamycin-resistant T2 lines were generated from hemizygous T1 plants, giving a kanamycin resistance : sensitive segregation ratio of 3 : 1. Seven Col-0 genomic sequences (described in Fig. S2) were subcloned into the multicloning site of binary vector pPZP212 (Hajdukiewicz et al., 1994). Six sequences were restriction fragments of BAC Col-0 genomic clones MSL1 and MVC8. The sequence of At3g16030 (tRFO3) was PCR-amplified from Col-0 DNA with forward primer 5′-gcaggtaccagtcagagtgatttttccgc-3′ and reverse primer 5′-gcaggtacctcaaaacgattgattcgaacc-3′ and subcloned into the KpnI site of pPZP212. To create the promoter–reporter gene fusion RFO3p:uidA, a 976 bp intergenic sequence was PCR-amplified using forward primer 5′-cggaattctagttgtttttgatgaagacaa-3′ complementary to sequence downstream of the stop codon of At3g16020 and reverse primer 5′-cggaattcaatttcagattttctgaaacttg-3′ complementary to sequence upstream of the start codon of At3g16030, digested with EcoRI and then subcloned into the EcoRI site of binary vector pORE R2 (Coutu et al., 2007).
DNA markers were PCR-amplified from crude leaf preparations and analyzed as in Celenza et al. (1995). Presence (rfo3) or absence (RFO3) of Sail_1212_G06 T-DNA insertion in At3g16030 was detected by PCR with the left primer 5′-taacaatccagttttgggacg-3′ and Sail-LB1 5′-gccttttcagaaatggataaatagccttgcttcc-3′ or the right primer 5′-gaaaagccacccttaaaccag-3′, respectively. Progeny of (rfo3 x Ty-0) x Ty-0 and (RFO3 x Ty-0) x Ty-0 were genotyped using dominant multiplex PCR markers CHR1.8, CHR1.2 and CHR3.3 that coamplify, respectively, RFO1-, RFO2- and RFO3-linked sequences of different lengths from Col-0 and no sequence from Ty-0 (Qiagen Multiplex PCR kit #206143). See Table S2 for primer sequences for multiplex PCR primers.
RFO3-T sequence (Genbank JN584215) was a contig assembled from PCR sequencing. Both strands of three overlapping PCR products amplified from Ty-0 DNA were sequenced. Descriptions of primers and PCR products are given in Table S3. cDNAs of RFO3-C and RFO-Tb-related transcripts were PCR-amplified from Col-0 and Ler cDNA libraries using forward primer 5′-gtggatccatgtggtcaaattgcatctttc-3′ and reverse primer 5′-gtggtacctcatcttgcttccatcactgtg-3′, and contigs were assembled from PCR sequencing of both cDNA strands (Minet et al., 1992; Kieber et al., 1993). Positions of introns in RFO3-C and RFO3-Tb were gaps in alignment of cDNA and genomic sequences. Augustus gene prediction software (http://augustus.gobics.de/) identified exon/intron boundaries in RFO3-Ta that were equivalent to those in RFO3-C. Transcripts of RFO3-C, RFO3-Ta and RFO3-Tb in total RNA, isolated using RNAeasy Plant Mini Kit (Qiagen), were detected as PCR-amplified cDNA, synthesized using SuperScript III (Invitrogen), using gene-specific primers (Table S4).
Observing glycosidase activity
Harvesting, cleaning and staining of roots are described in Diener (2012). To visualize FOM infection, roots were incubated in 30-fold excess staining buffer with 0.02% 5-bromo-4-chloro-3-indoxyl (X)-α-l-arabinofuranoside (X-Ara, Gold Biotechnologies Inc., St Louis, MO, USA) at 28°C overnight. To visualize RFO3p:uidA expression, cleaned roots were incubated in staining buffer with 0.02% X-β-d-glucuronide (X-Gluc) at 37°C overnight. To quantify FOM-derived arabinofuranosidase (ABF) activity, freshly harvested roots were incubated for 16 h at 28°C in 30-fold excess staining solution with 0.04% 4-nitrophenyl-α-l-arabinofuranoside (NP-Ara). Relative ABF activity, as OD410 nm g–1 FW root, was estimated by subtracting OD600 from OD410.
Nucleotide and amino acid sequences were from the Phytozome database v8.0 (www.phytozome.net). Unambiguously best-matched sequences to RFO3/At3g16030 or its paralog At1 g67520 in The Arabidopsis Information Resource database (TAIR, www.arabidopsis.org), according to Blastp, were homologs of RFO3 (Table S5).
Sequence information for 217 accessions listed in Table S6 were downloaded from the Salk Institute 1001 Genomes Project website (http://signal.salk.edu/atg1001). Data files were ‘quality_variant_filtered_<strain_name>.txt’ and ‘unsequenced_<strain_name>.txt’. Gene annotation for five Arabidopsis chromosomes was downloaded from the TAIR website. A detailed summary of single nucleotide polymorphisms (SNPs) in 17 RFO3-C sequences from 102 accessions is given in Table S7. Mean HI values of 53 tested accessions are given in Table S8. HI scores of four to seven FOM-infected plants were evaluated at 16 and 20 d post-soil inoculation (dpi). For ranking, mean HI values at 20 dpi had priority over values at 16 dpi, and accessions with identical values were assigned the same intermediate rank. SNPs were common when variant alleles were present in more than five tested accessions and fewer than 49 accessions. For testing association, rank distributions of accessions with the reference and substitution bases were compared using the Mann–Whitney U-test. Descriptions of the 25 SNPs with the most extreme positive and negative Z-values are given in Table S9.
Mapping quantitative resistance
Previously, the quantitative resistance trait RFO3 was detected on chromosome 3 in A. thaliana (Diener & Ausubel, 2005). We confirm this QTL by examining the segregation of RFO3 in the absence of other major RFO QTLs. The recombinant plant 6E5 from an earlier study was heterozygous for Col-0 and Ty-0 alleles of RFO3 (RFO3-C/T) and homozygous for the susceptible Ty-0 allele at other RFO loci (according to DNA markers linked to six RFO QTLs; Diener & Ausubel, 2005). Among self progeny of 6E5, the resistant allele of RFO3 from Col-0 (RFO3-C) and the susceptible allele from Ty-0 (RFO3-T) expressed incomplete dominance: RFO3-C/T heterozygotes were more susceptible than RFO3-C homozygotes (RFO3-C/C) and more resistant than RFO3-T homozygotes (RFO3-T/T, Fig. 1a).
To identify the gene responsible for quantitative resistance, we used the segregation of RFO3 to map the resistance trait to a genetic interval corresponding to 89 kb (see the 'Materials and Methods' section for a description of linkage mapping). In order to map RFO3, we needed to reliably assign RFO3 genotypes to individual plants. However, the relative resistance of a plant could not reliably report a plant's genotype because RFO3 has a quantitative effect and symptoms vary even in plants with the same genotype. Thus, the RFO3 genotype of a plant was indirectly ascertained by testing whether or not the resistance and genotype of an RFO3-linked marker showed coinheritance in a sample of the plant's progeny. The RFO3 genotypes of selected F1BC plants (from a previously described mapping population) were determined in tests of self (F2BC) progeny (Diener & Ausubel, 2005). The position of recombination breakpoints relative to RFO3 in these lines placed RFO3 between DNA markers CIW11 and NGA162. This map position was then refined with new recombination breakpoints between CIW11 and NGA162 in the tested F2BC. In particular, two F2BC plants, 1E2.A6 and 2C3.B9, which proved to be RFO3-C/T, had tightly linked recombination breakpoints on either side of RFO3 that were no more than 220 kb apart (between markers MJK13 and MSL1.60 in Fig. 1b). Screening of more recombinant plants identified a recombination breakpoint between markers MVC8 and MSL1.29 in plant 1D3 (Fig. 1b). Analysis of 1D3′s self progeny indicated that 1D3 was RFO3-C/T and placed RFO3 between markers MVC8 and MSL1.60 and in an interval with 27 candidate genes (Fig. S2).
An RLK gene confers quantitative resistance
One candidate gene, At3g16030, enhanced resistance to FOM as a transgene. In total, seven Col-0 genomic subclones, representing 72% of the final RFO3 interval and including 13 candidate gene sequences, were stably introduced to Ty-0 as transgenes (Fig. S2). Independent plants from transformation using the full-length genomic sequence of At3g16030 and no other gene (tRFO3) reproducibly expressed more resistance to FOM than untransformed Ty-0 (Fig. 1c); and, a pure-breeding line made homozygous for tRFO3 expressed resistance to FOM that was intermediate to Ty-0 and Col-0 (Fig. 1c). On the other hand, independent plants from transformation using the six other genomic clones without At3g16030 exhibited no more resistance to FOM than untransformed Ty-0 (Fig. S2).
Disruption of At3g16030 by T-DNA insertion enhanced susceptibility to FOM. Homozygous lines in the Col-0 genetic background with (rfo3) and without (RFO3) T-DNA insertion in At3g16030 were isolated from a seed pool segregating for the sequence-mapped insertion Sail_1212_G06. According to the PCR sequence across the insertion site, the T-DNA left border sequence interrupts the coding sequence 379 bp downstream of the start codon and in the first exon. When infected with FOM, rfo3 was marginally less resistant than wild-type Col-0, even though rfo1, which is also in the Col-0 genetic background, clearly expressed visible disease symptoms in the same assay (Fig. 2a). Presumably, the remaining RFO QTLs in rfo3, and especially RFO1, expressed sufficient resistance in the absence of RFO3. When rfo3 was combined with rfo1, the double mutant rfo1 rfo3 exhibited more severe symptoms than rfo1 alone (Fig. 2a).
At3g16030 alone accounted for the QTL as rfo3 abolished quantitative resistance at RFO3. Because RFO3 was detected in the cross (Col-0 × Ty-0) x Ty-0 (Diener & Ausubel, 2005), we re-examined resistance in two new crosses that recreated the original cross, with the exception that the isogenic RFO3 (wildtype) or rfo3 lines from Sail_1212_G06 replaced Col-0 in the parental cross to Ty-0. Among progeny of the new crosses, (RFO3 × Ty-0) × Ty-0 and (rfo3 × Ty-0) × Ty-0, stronger resistance was observed among Col-0/Ty-0 heterozygotes than Ty-0 homozygotes at RFO1 and RFO2 (Fig. 2b), and these results reproduced the previous detection of RFO1 and RFO2. However, progeny of the crosses yielded different results at RFO3. Among progeny of the RFO3 parent, Col-0/Ty-0 heterozygotes again exhibited more resistance than Ty-0/Ty-0, whereas wilt disease similarly affected the two genotypes among progeny of the rfo3 parent (Fig. 2b). Thus, At3g16030, which encodes an S domain subfamily 1 (SD1) RLK, alone conferred quantitative resistance.
Susceptible allele is highly diverged
Sequence of the susceptible allele of RFO3 in Ty-0, which encodes two At3g16030-related SD1 RLK genes, has diverged remarkably from the Col-0 sequence. To appreciate what distinguishes resistance from susceptibility at RFO3, we assembled a 7526 bp sequence spanning the RFO3 locus in Ty-0. This RFO3-T sequence had the highest identity to the shorter 3860 bp sequence spanning the RFO3 sequence in the reference Col-0 genome (Fig. 3a). Two SD1 RLK genes in RFO3-T, RFO3-Ta and RFO3-Tb are separated by 391 bp and oriented head to tail in the same direction as the single SD1 RLK gene RFO3-C. The transcript of RFO3-C was detected using RT-PCR in Col-0 RNA from leaves and roots and not in Ty-0 RNA (Fig. S3a,b); transcripts of RFO3-Ta and RFO3-Tb were detected in Ty-0 RNA from leaves and roots and not in Col-0 RNA (Fig. S3a,b); and transcripts of all three RFO3 genes were similarly detected in RNA from an equal mixture of Col-0 and Ty-0 leaves (Fig. S3c). As shown in Fig. 3a, the RFO3-C sequence aligned to both RFO3-Ta and RFO3-Tb with similar overall identity, suggesting that the two SD1 RLK genes in RFO3-T originated from duplication of an ancestral RFO3 gene (Fig. 3a). In the alignments of RFO3-Ta, RFO3-Tb and RFO3-C, a higher nucleotide identity in exons coding for the conserved protein kinase domain (c. 98%) than in the intervening introns (c. 80%, in Fig. 3a,b) provided evidence of purifying selection and suggested that the two SD1 RLK genes in RFO3-T expressed function for at least some time after duplication.
Homologous receptor genes at RFO3
Transcripts of the RFO3 genes encode comparable full-length proteins with 871 (RFO3-C), 852 (RFO3-Ta) and 881 (RFO3-Tb) residues. In pairwise comparisons, the amino acid sequences are between 81 and 87% identical. The protein sequence of RFO3 RLKs can be divided into six domains previously defined in related SD1 RLKs (see translated domains of RFO3-C in Fig. S4): An initial signal peptide (1) would target the following three domains to extracellular space: the mature amino-terminal B-lectin domain (2), the S-domain (3) and the membrane-proximal PAN-APPLE domain (4). The transmembrane domain (5) would anchor the protein in the cell membrane and target the carboxy-terminal kinase domain (6) to the cytoplasmic side of the membrane (Naithani et al., 2007).
According to bioinformatic analysis, RFO3 belongs to a gene lineage that is not conserved in plant species outside the mustard family. RFO3 and its paralog (At1 g67520) are in a distinct gene lineage in the larger SD1 subfamily of RLK genes in A. thaliana (Shiu et al., 2004). Unambiguous orthologs and paralogs of RFO3 were identified in a subset of related crucifer species and no other taxonomic families (Table S5) in the Phytozome v8.0 plant genome database. In crucifers Arabidopsis lyrata, Capsella rubella and Brassica rapa, single homologs are in chromosomal regions with similar gene organization as RFO3 in A. thaliana (Fig. S5). However, in crucifer Thellungiella halophila, neither RFO3 nor At1 g67520 was the best-matched A. thaliana gene for an SD1 RLK homolog, and no RFO3-related sequence is present in the chromosomal region of T. halophila corresponding to the RFO3 locus.
Among RFO3 orthologs and even alleles, there is considerable divergence in sequence coding for extracellular domains and especially the S domain (Fig. S6). In particular, three regions in the amino acid sequence of S domains of RFO3-C, RFO3-Ta and RFO3-Tb (as well as RFO3 orthologs) were poorly aligned. Two of the three regions in RFO3 homologs were at positions comparable to two so-called hypervariable regions (hvII and hvIII) in the S domain of Brassica SRK (Kusaba et al., 1997), and the third region in RFO3 homologs was roughly 30 amino acids removed from a third hypervariable region (hvI) in SRK.
Vascular-expressed RFO3 restricts root infection
To address where in the plant RFO3 affects quantitative resistance to FOM, we visualized expression of RFO3 in whole plants using the reporter gene RFO3p:uidA, which fuses the promoter of RFO3-C to the β-glucuronidase (GUS) gene uidA. In independent transformed lines of RFO3p:uidA, X-Gluc staining for GUS activity was strongest in rosette leaves, where symptoms occur during the expressive phase of infection, and was especially intense in vascular tissue of leaves and the root system, which is colonized by F. oxysporum during the secondary determinative phase (Fig. 4a). Less intense staining appeared in young leaves, flowers and root tips, and the relative intensity of staining was consistent with a report of higher to lower accumulation of RFO3 transcript in leaves, roots and flowers (Niwa et al., 2006). Expression of the RFO3 promoter, being strong in vascular tissue throughout the plant and especially in leaves, did not hint at whether resistance was a consequence of reduced fungal infection in roots or reduced symptoms in shoots.
To resolve whether root and/or shoot expression of RFO3-C affected resistance, disease progression was compared in grafted plants expressing RFO3-C in shoots and roots, shoots alone, roots alone or neither roots nor shoots. The four combinations of root and shoot genotypes were created by all possible grafts between rootstocks and scions of Ty-0 (RFO3-T/T) and the near-isogenic line 1E2 (RFO3-C/C). Line 1E2 was a product of mapping RFO3 and represents the introgression of RFO3-C (and a segment of the Col-0 chromosome 3, between markers CIW11 and NGA162) into the Ty-0 genetic background that lacks the resistance provided by five other RFO QTLs (Diener & Ausubel, 2005). The disease progression in grafted plants with RFO3-T in both roots and shoots was similarly observed in grafted plants that expressed RFO3-C in shoots and RFO3-T in roots and showed that strong shoot expression of RFO3-C was inconsequential for resistance (Fig. 4b). Meanwhile, the enhanced resistance of plants with RFO3-C in roots and shoots was comparable in plants with RFO3-T in shoots and RFO3-C in roots (Fig. 4b). Thus, only root-expressed RFO3-C enhanced resistance.
By examining F. oxysporum infection in roots, we found that resistance correlated with reduced infection of the root system. In theory, root-expressed RFO3-C might instead suppress symptoms at a distance in the shoot. At 10 dpi, when early symptoms, such as stunted growth and darkening of leaves, were more modest in RFO3-C/C (1E2) than in RFO3-T/T (Ty-0, Fig. 5a), FOM infection was visualized and quantified using colorimetric reagents that are hydrolyzed by a Fusarium-derived ABF activity (Diener, 2012). Showing a clear difference in FOM infection, the indigogenic substrate 5-bromo-4-chloro-3-indoxyl-α-l-arabinofuranoside (X-Ara) produced stronger and more prevalent blue staining in whole roots of RFO3-T/T than of RFO3-C/C (Fig. 5b). Indeed, RFO3-T/T accumulated 3.5-fold more Fusarium-derived ABF activity than RFO3-C/C, as detected by conversion of colorless 4-nitrophenyl-α-l-arabinofuranoside (NP-Ara) to soluble yellow product, 4-nitrophenol (Fig. 5c).
Quantitative resistance is specific to FOM
Line 1E2 (RFO3-C/C), which has RFO3-C but is otherwise the Ty-0 genetic background, and Ty-0 (RFO3-T/T) expressed similar resistance to crucifer-infecting formae speciales other than FOM. In Fig. 6(a), both RFO3-T/T and RFO3-C/C had incomplete resistance to FOC race 1 and FOR, and were susceptible to FOC race 2. Only when infected with FOM did RFO3-C/C express more resistance than RFO3-T/T (Fig. 6a). In theory, both RFO3-C/C and RFO3-T/T might, in fact, express the same resistance to FOC if RFO3-C and RFO3-T were functionally equivalent during infection with FOC. However, rfo3, which is deficient for RFO3-C in the Col-0 genetic background, was no more susceptible than wildtype Col-0 (RFO3-C/C) in the same infection assay in which rfo1 expressed enhanced susceptibility (Fig. 6b). Furthermore, when combined with rfo1, rfo3 failed to appreciably affect the susceptibility of rfo1 in the double mutant rfo1 rfo3 (Fig. 6b).
RFO3-C conferred resistance to independent isolates of FOM race 2. Although disease progressed faster (with NRRL 22545) or slower (with NRRL 38343) depending on the FOM race 2 isolate used for infection, RFO3-C/C quantitatively expressed more resistance to both isolates than RFO3-T/T (Fig. 6c). Also, resistance to F. oxysporum in the Ty-0 genetic background is not generally compromised, as both Ty-0 and 1E2 manifested complete resistance to two FOM race 1 isolates (Fig. 6c).
RFO3-C and RFO3-T represent ancestral variation
Because RFO3 is a natural resistance trait, we investigated the diversity of its sequence in 217 wild accessions of A. thaliana and found that RFO3-C and RFO3-T are representative of two ancestral variants of RFO3, each of which is present in roughly one-half of available accessions. The 1001 Genomes Project, using the existing Col-0 reference genome to guide assembly (or resequencing) of whole genome sequence, has made available two types of sequence information: for the accession sequence that aligns to the reference genome, there is an index of SNPs with corresponding bases from reference and accession sequences; and for the reference sequence that fails to align to the accession sequence (because the reference sequence is absent or highly diverged in accession DNA) there is an index of unsequenced gaps in the reference sequence (Ossowski et al., 2008).
In the 53% of accessions with an RFO3-T-like sequence, unsequenced gaps in RFO3 corresponded to the highly diverged sequence in RFO3-C and RFO3-T. Also, because reference-guided resequencing would align both RFO3-Ta and RFO3-Tb sequences to the single RFO3-C sequence, two genes were not evident in resequencing results for Ty-0 (or other accessions). Nevertheless, 90% of the 66 common SNPs in partially resequenced RFO3 sequences were also polymorphisms in the alignment of RFO3-T and RFO3-C, indicating that RFO3-T was representative of these partial RFO3 sequences.
In the remaining 47% of resequenced accessions, RFO3 was not less than 99.5% identical to RFO3-C. Among the 34 SNPs that differentiated these RFO3-C-like sequences, 10 were deemed to be common SNPs, because the alternative bases were present in three or more of the 17 RFO3-C-like sequences (Table S7). Tight clustering of six common SNPs in a 35 bp sequence, coding for the extracellular B lectin domain, was conspicuous and contrasted with the more haphazard distribution of 24 rare SNPs across the length of RFO3 (Fig. S4). Extensive recombination among the six clustered SNPs in RFO3-C-like sequences suggests that this variation has been preserved and possibly selected.
According to genotypes at common SNPs, RFO3-C-like sequences were group into three subtypes, namely C1, C2 and C3. Sixty per cent of accessions had the C1 subtype and their RFO3 sequences were identical to RFO3-C (with the exception of a single unique SNP in Er-0). Absence of diversity in this majority of RFO3-C-like sequences suggested that representation of the C1 subtype had recently expanded in wild populations. Possibly favoring the C1 subtype, one SNP, conditioning a nonsynonymous codon substitution, from arginine to methionine in the S domain, distinguished C1 from C2 and C3 subtypes (Fig. S6, Table S7). Twelve per cent of accessions had the C2 subtype and the same genotype at common SNPs, although expansion of this haplotype in the population appeared older than expansion of the C1 subtype, because 13 rare SNPs have accumulated in the five C2 sequences. The remaining 28% of accessions with the C3 subtype have an assortment of minor genotypes with recombined base combinations at nine common SNPs. One common SNP, conditioning either lysine or glutamic acid in the S domain, distinguished both C1 and C2 subtypes from the C3 subtype (Fig. S6, Table S7). Finally, just one SNP, in 2% of all accessions, was expected to have an obvious deleterious effect, a frameshift in protein translation.
Similar diversity at common SNPs surrounding RFO3-C- and RFO3-T-like sequences supported the ancestral origin of these two highly diverged variants (Fig. S7). In theory, genotypes at RFO3 and any common SNP, even SNPs that are proximal to RFO3, should show independent assortment among resequenced accessions unless historical recombination has had insufficient time to disassociate any original linkage between RFO3 and the SNP. For example, the absence of diversity in sequences neighboring RFO3 in accessions with the C1 and C2 subtypes supported the recent expansions (and possibly partial selective sweeps) by C2 and, more recently, C1 subtypes (Fig. S7). Interestingly, RFO3 variants and subtypes showed broad and overlapping geographical distributions when accessions were placed on a map (Fig. S8). In particular, the relatively young C1 subtype appeared to have spread to most regions sampled by the resequenced accessions.
RFO3 is major determinant of resistance
Testing the association of resistance with common SNPs genome-wide, we found that the ancestral RFO3-C-like variant is a major determinant of resistance to FOM in wild populations. Because Arabidopsis accessions express considerable and continuous variation in symptom severity, an arbitrary sample of 53 accessions could be ordered from the least to the most resistant. The association of 593 091 common SNPs within 33 325 annotated genes, including 67 SNPs in RFO3, with resistance was evaluated, and the difference of resistance expressed by accessions with and without the substitution base at a SNP was expressed as a Z-statistic. Positive or negative values of Z correlated resistance with the reference or substitution base, respectively. The values of Z for SNPs in RFO3 and genes surrounding RFO3 on chromosome 3 are depicted in Fig. 7. Remarkably, among the 25 SNPs with the most extreme positive Z, 19 were located within RFO3 or TMA7 (the gene immediately downstream of RFO3); no other gene in Arabidopsis had such a preponderance of SNPs with extreme (positive or negative) values of Z (Table S9).
In separate comparisons, accessions with each of the three RFO3-C-like subtypes were more resistant than accessions with the RFO3-T-like variant (Fig. 8). Clearly, the C1 subtype that was identical to the reference genome was associated with resistance. However, correlation of C2 and C3 subtypes with resistance was also evident in the few SNPs in RFO3 with substantial negative Z values (in Fig. 7), representing common SNPs that distinguish C2 and C3 subtypes from the C1 subtype (Col-0) and the RFO3-T-like variant.
RFO3 affects quantitative resistance much as Immunity (I) traits in tomato affect monogenic resistance to F. oxysporum f.sp. lycopersici (FOL). Wilt resistance conferred by RFO3-C in A. thaliana and I genes in tomato is localized to the root system and corresponds to less F. oxysporum infection in root vascular tissue (Beckman & Roberts, 1995). Strong expression of RFO3 and Immunity-2 (I-2) in vascular tissue surrounding xylem vessels suggests that these genes function in local immune response (Mes et al., 2000). Also, RFO3-C and I genes provide resistance to specific F. oxysporum pathogens, and this specificity suggests that RFO3-C, like I genes, is a participant in gene-for-gene resistance (Takken & Rep, 2010). RFO3-C enhanced resistance to FOM but had no effect on resistance to FOC and FOR, just as I-2 confers strong resistance to FOL race 2 but only partial or no resistance to FOL races 1 and 3, respectively. Conforming to the gene-for-gene concept, I-2 recognized FOL race 2, because FOL race 2 expresses Avirulence 2 (Avr2) (Houterman et al., 2009). At a molecular level, the NB-LRR protein I-2 perceives the small cysteine-rich effector protein Avr2 in host plant cytoplasm (Houterman et al., 2009). Because RFO3 encodes RLKs, RFO3-C more likely functions as a PRR and perceives an extracellular Fusarium-derived signal that is present in FOM infection and absent in FOC and FOR infections (Thomma et al., 2011).
The coincidence that RFO3 and the previously published RFO1 are RLK genes shows that diversity in RLK genes can be a major source of quantitative resistance in wild plant populations such as A. thaliana. By contrast, strong monogenic resistance and the hypersensitive response are largely but not exclusively associated with diversity in NB-LRR genes (McHale et al., 2006). Interestingly, a recent genome-wide analysis of quantitative variation in resistance to northern leaf blight in maize identified several candidate SNPs associated with RLK genes and none with NB-LRR genes (Poland et al., 2011).
In fact, plant genomes have sufficient numbers of candidate PRR RLK genes to support their ubiquitous role in polygenic resistance traits. For example, 233 and 639 genes in the reference genomes of Arabidopsis and rice, respectively, belong to nine RLK gene subfamilies (DUF26, L-LEC, LRK10L-2, LRR-I, LRR-VIII-2, LRR-XII, SD1, SD2b, and WAK) whose history, genome organization and expression implicate their involvement in pathogen recognition (Lehti-Shiu et al., 2009). These RLK gene subfamilies are especially up-regulated by biotic stress conditions, are overrepresented in tandem gene clusters and have extensive lineage-specific expansion in Arabidopsis, poplar, and rice genomes. In these gene families, orthologous sequences are often highly diverged and duplicated or lost in other plant genomes (Lehti-Shiu & Shiu, 2012). Indeed, genes at RFO3 belong to a gene lineage that is present in only a subset of species in the mustard family.
The highly diverged, ancestral variation at RFO3 is unusual for a putative PRR resistance trait. Previous examples of resistance genes encoding RLKs or receptor-like proteins (RLPs) come from domesticated species, and most of these lack substantial intraspecific diversity. For instance, susceptible alleles of Ve1 and Pi-d2 associate single SNPs with loss of resistance and introduce, respectively, a premature stop codon and an amino acid substitution in the transmembrane domain (Chen et al., 2006; Fradin et al., 2009). The presence of Cf genes in tomato and Xa21 in rice is a consequence of interspecific breeding, as these traits were absent in the cultivated species (Wang et al., 1996; Thomas et al., 1998). Only at the rice Xa26/Xa3 locus are both resistant and susceptible alleles represented by diverged RLK homologs in cultivated rice (Sun et al., 2006). Possibly, domestication has greatly diminished the functional diversity of PRR genes that normally exists in wild species populations, and variation that now exists in cultivated species largely arises from loss-of-function polymorphisms.
An intriguing aspect of RFO3′s identity is homology to SRK, which is the defining member of the SD1 RLK gene family (Stein et al., 1991). SRK has no known role in defense response, although pathogen infection strongly induces expression of its paralog ARK3 in A. thaliana (Pastuglia et al., 2002; Tantikanjana et al., 2010). Rather, SRK is involved in the self-incompatibility (SI) response, a developmental process in outcrossing species that inhibits self-fertilization and consequently prevents inbreeding depression (Hatakeyama et al., 2000). Because dissimilar mechanisms govern SI in different plants species, SI appears to have multiple evolutionary origins in plant phylogeny. The SI response of Brassicaceae, in particular, is proposed to be an adaptation of a defense response because there are striking parallels in SRK-mediated SI and immunity (Sanabria et al., 2008). Comparisons with how SRK recognizes self could give insights into how RFO3-C recognizes nonself. As the female determinant in SI, the plasma membrane-spanning receptor SRK is expressed at the stigma surface where pollen is received (Takasaki et al., 2000). SRK's ligand, called S-locus cysteine-rich protein (SCR), is the male determinant and decorates the outer pollen coat (Kachroo et al., 2001).
Considering the perception of SCR by SRK, we imagine that RFO3-C might be a receptor for a FOM-derived protein just as SRK is the receptor of SCR. In fact, SCR has physical properties similar to secreted effector proteins produced by pathogenic fungi, including F. oxysporum (Rafiqi et al., 2012). Several SECRETED IN XYLEM (SIX) genes of FOL express small cysteine-rich proteins, including the three known avirulence factors recognized by four I genes (Houterman et al., 2007; Takken & Rep, 2010).
On the other hand, RFO3-C's perception of an FOM-derived signal might be indirect and mediated by interaction with an SCR-related protein. In tomato, for example, the PRR Cf-2 indirectly perceives Avr2, an avirulence protein from fungal pathogen Cladosporium fulvum, through interaction with Rcr3 (Rooney et al., 2005). If interaction between SCR and SRK predates the SI response in Brassicaceae, ancestors of both SCR and SRK might function together in pathogen recognition. The FOM-derived signal might target and activate one or more SCR-like host proteins. The 29 SCR-like genes in the Arabidopsis reference genome are members of a larger defensin-like (DEFL) gene family with over 300 members (Vanoosthuyse et al., 2001; Silverstein et al., 2005).
Resemblance of sequence divergence in RFO3 variants and SRK alleles suggests that both RFO3 variants express functional PRRs with distinct ligand specificities. Highly diverged SRK alleles express homologs that function in the same SI response but perceive different SCR ligands. Likewise, the RFO3-T allele encodes two full-length homologs of RFO3-C without obvious defects. If RFO3-T were simply nonfunctional as a result of a recent or rare loss-of-function polymorphism, resistance and susceptibility would not be associated with the RFO3-C-like and RFO3-T-like variants, respectively. Dissimilarity in RFO3 homologs is concentrated in the extracellular S domain, the apparent binding domain for SCR in SRK homologs (Boggs et al., 2009). Extreme sequence divergence in RFO3 homologs roughly coincides with the hypervariable regions that are especially critical for haplotype-specific activation of SRK.
We studied RFO3 in an experimental pathosystem in a controlled laboratory environment, which begs the question: is RFO3, as named, responsible for resistance to FOM in the real world? Or consider the corollary: is FOM a real pathogen of A. thaliana? We are aware of FOM only because it is a problem for commercial growers of the ornamental plant M. incana. FOM is isolated from diseased garden stock in North America, continental Europe, the UK and Japan, so FOM appears to have a wide geographical range (O'Donnell et al., 2009), and this makes a chance encounter of FOM and widely dispersed A. thaliana plausible (Mitchell-Olds, 2001). FOM corresponds to a single clonal lineage in the F. oxysporum species complex, although varieties of M. incana can differentiate two races (Bosland & Williams, 1987; O'Donnell et al., 2009). Similarly, we observed only complete resistance to FOM race 1, even in Arabidopsis accessions that are susceptible to FOM race 2 (as shown for Ty-0 in Fig. 6c). Because we observed enhanced resistance to three independent isolates of FOM race 2, RFO3-C probably confers resistance to FOM in general. In some instances, the appearance of exclusive one-to-one interactions between particular resistance genes and particular avirulence genes (from particular pathogens) may be an artifact of classical gene-for-gene analysis (de Jonge et al., 2012; Lozano-Torres et al., 2012). RFO3-C might confer resistance to other pathogens as well, which could explain our evidence for recent partial selective sweeps of RFO3 subtypes C1 and C2 that provide similar resistance to FOM. Thus, in the field, RFO3 possibly confers resistance to FOM as well as other potential pests and pathogens.
We thank the Dean of Life Sciences, University of California, Los Angeles, for research funding. S.J.C. was supported by a NIH Genetics Training Grant to UCLA.