(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



A new family of structurally conserved fungal effectors displays epistatic interactions with plant resistance proteins [1]

['Noureddine Lazar', 'Institute For Integrative Biology Of The Cell', 'Université Paris-Saclay', 'Cea', 'Cnrs', 'Gif-Sur-Yvette', 'Carl H. Mesarich', 'Laboratory Of Molecular Plant Pathology', 'School Of Agriculture', 'Environment']

Date: 2022-09

Recognition of a pathogen avirulence (AVR) effector protein by a cognate plant resistance (R) protein triggers a set of immune responses that render the plant resistant. Pathogens can escape this so-called Effector-Triggered Immunity (ETI) by different mechanisms including the deletion or loss-of-function mutation of the AVR gene, the incorporation of point mutations that allow recognition to be evaded while maintaining virulence function, and the acquisition of new effectors that suppress AVR recognition. The Dothideomycete Leptosphaeria maculans, causal agent of oilseed rape stem canker, is one of the few fungal pathogens where suppression of ETI by an AVR effector has been demonstrated. Indeed, AvrLm4-7 suppresses Rlm3- and Rlm9-mediated resistance triggered by AvrLm3 and AvrLm5-9, respectively. The presence of AvrLm4-7 does not impede AvrLm3 and AvrLm5-9 expression, and the three AVR proteins do not appear to physically interact. To decipher the epistatic interaction between these L. maculans AVR effectors, we determined the crystal structure of AvrLm5-9 and obtained a 3D model of AvrLm3, based on the crystal structure of Ecp11-1, a homologous AVR effector candidate from Fulvia fulva. Despite a lack of sequence similarity, AvrLm5-9 and AvrLm3 are structural analogues of AvrLm4-7 (structure previously characterized). Structure-informed sequence database searches identified a larger number of putative structural analogues among L. maculans effector candidates, including the AVR effector AvrLmS-Lep2, all produced during the early stages of oilseed rape infection, as well as among effector candidates from other phytopathogenic fungi. These structural analogues are named LARS (for Leptosphaeria AviRulence and Suppressing) effectors. Remarkably, transformants of L. maculans expressing one of these structural analogues, Ecp11-1, triggered oilseed rape immunity in several genotypes carrying Rlm3. Furthermore, this resistance could be suppressed by AvrLm4-7. These results suggest that Ecp11-1 shares a common activity with AvrLm3 within the host plant which is detected by Rlm3, or that the Ecp11-1 structure is sufficiently close to that of AvrLm3 to be recognized by Rlm3.

An efficient strategy to control fungal diseases in the field is genetic control using resistant crop cultivars. Crop resistance mainly relies on gene-for-gene relationships between plant resistance (R) genes and pathogen avirulence (AVR) genes, as defined by Flor in the 1940s. However, such gene-for-gene relationships can increase in complexity over the course of plant-pathogen co-evolution. Resistance against the plant-pathogenic fungus Leptosphaeria maculans by Brassica napus and other Brassica species relies on the recognition of effector (AVR) proteins by R proteins; however, L. maculans produces an effector that suppresses a subset of these specific resistances. Using a protein structure approach, we revealed structural analogy between several of the resistance-triggering effectors, the resistance-suppressing effector, and effectors from other plant-pathogenic species in the Dothideomycetes and Sordariomycetes classes, defining a new family of effectors called LARS. Notably, cross-species expression of one LARS effector from Fulvia fulva, a pathogen of tomato, in L. maculans resulted in recognition by resistant cultivars of oilseed rape. These results highlight the need to integrate knowledge on effector structures to improve resistance management and to develop broad-spectrum resistances for multi-pathogen control of diseases.

Funding: YPH was funded by a “Contrat Jeune Scientifique” grant from INRAE ( www.inrae.fr ) and NT by a PhD salary from the University Paris-Saclay ( www.universite-paris-saclay.fr ). The “Effectors and Pathogenesis of L. maculans” group benefits from the support of Saclay Plant Sciences-SPS (Agence Nationale de la Recherche grant ANR-17-EUR-0007; www.anr.fr ). This work was supported by the Agence Nationale de la Recherche projects StructuraLEP (ANR-14-CE19-0019 to IF and HVT) and Ln23 (ANR-13-BS07-0007-01 to JG), the Plant Health and Environment division of INRAE (Resistrans Project to IF), the Australian Grains Research and Development Corporation (UM00050 to AI; https://grdc.com.au/ ), French Infrastructure for Integrated Structural Biology (FRISBI) ANR-10-INBS-05, and by funds from the Centre National de la Recherche Scientifique to HVT ( www.cnrs.fr ) and the University Paris-Saclay to HVT. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Here we describe the 3D structures of AvrLm5-9 and AvrLm3, whose recognition by Rlm9 and Rlm3, respectively, is masked by the presence of AvrLm4-7. Surprisingly, despite low sequence similarity, AvrLm5-9 and AvrLm3 are structural analogues of AvrLm4-7, sharing an anti-parallel β-sheet covered by α-helices. Structure-informed and pattern-based searches identified a larger number of putative structural analogues among AVR effectors and effector candidates of L. maculans, but also of other phytopathogenic fungi, including Ecp11-1 from the biotrophic tomato leaf mold fungus Fulvia fulva (formerly Cladosporium fulvum). Remarkably, transformants of L. maculans producing F. fulva Ecp11-1 triggered Rlm3-mediated immunity and this resistance could be suppressed by AvrLm4-7. These findings will enable hypotheses to be made about the way effectors suppress ETI and can guide recommendations on how to use plant R genes targeting AVRs belonging to structural families of effectors.

Elucidation of the 3D structures of effectors may provide an effective strategy to resolve functional traits. Indeed, structure determination of effectors and the proteins with which they interact has provided key advances in our understanding of plant-pathogen interactions, including: the identification of protein functions that were not apparent from sequence analysis alone, the visualization of molecular interfaces of relevance to pathogen virulence and to plant immunity, and the identification of structural homologies in effectors that were not visible by sequence comparisons (reviewed in [ 23 ]). The crystal structure of AvrLm4-7 did not reveal similarities with documented effectors, but suggested a positively charged surface patch could be involved in AvrLm4-7 translocation into the cytoplasm of plant cells [ 24 ]. AvrLm4-7 escapes Rlm4-mediated recognition through a single point mutation [ 18 ] and Rlm7-mediated recognition through more drastic DNA changes (gene deletion, accumulation of mutations) or point mutations [ 25 ]. Blondeau et al. [ 24 ] identified a protein region involved in Rlm4-mediated recognition, and two regions involved in Rlm7-mediated recognition.

The Dothideomycete Leptosphaeria maculans, causal agent of oilseed rape stem canker or blackleg disease, is one of the fungal pathogens in which suppression of ETI by the presence of an AVR gene has been demonstrated. L. maculans can be controlled by combining qualitative and quantitative resistance of the host plant [ 12 ]. To date, ten AVR genes (called AvrLm) recognized by the products of R genes (called Rlm) from Brassica napus or other Brassica species have been identified [ 13 – 17 ] and share common characteristics: they encode small secreted proteins with no or low homologies in sequence databases, are located in repeat-rich regions of the genome, and are specifically expressed during the early stages of leaf infection. Among them, AvrLm4-7 suppresses Rlm3-mediated resistance triggered by AvrLm3 and Rlm9-mediated resistance triggered by AvrLm5-9 [ 9 , 14 ]. How AvrLm4-7 suppresses Rlm9- and Rlm3-mediated disease resistance is not known: the presence of AvrLm4-7 does not impede AvrLm3 and AvrLm5-9 expression, and yeast two-hybrid (Y2H) assays suggest the absence of a physical interaction between AvrLm4-7, AvrLm5-9 and AvrLm3. While AvrLm5-9 and AvrLm3 share 29% amino acid sequence identity, a very low level of identity was found with AvrLm4-7 (15%). AvrLm4-7 confers a dual recognition specificity by two distinct R proteins of oilseed rape, Rlm4 and Rlm7 [ 18 ], and loss of AvrLm4-7 is associated with a fitness cost [ 19 , 20 ]. AvrLm5-9 and AvrLm3, on the other hand, are always present in L. maculans isolates, and only point mutation polymorphisms were reported, suggesting a high importance of these two effectors in pathogenicity towards B. napus [ 14 , 21 , 22 ]. Moreover, the silencing of AvrLm3 led to a reduced aggressiveness [ 9 ].

Fungi are the most devastating pathogens of plants, including crops of major economic importance. They represent a recurrent threat to agriculture and possess extreme adaptive abilities, resulting in constant disease outbreaks [ 1 , 2 ]. Host invasion relies on effectors, key elements of pathogenesis, that modulate plant immunity and facilitate infection [ 3 , 4 ]. Fungal effector genes are diverse and typically encode small proteins, predicted to be secreted, with no or few homologues present in sequence databases, and an absence of known sequence motifs. In most phytopathogenic fungi, no large effector gene families have been identified [ 5 ]. Notably, effectors can have a dual role in plant-pathogen interactions, both targeting plant components and being targeted by plant resistance (R) proteins. Such dual-role effectors are known as avirulence (AVR) proteins because, in the presence of a corresponding R protein, they render the pathogen that produces them avirulent. Recognition of a pathogen AVR protein triggers a set of immune responses grouped under the term Effector-Triggered Immunity (ETI), frequently leading to a rapid localized cell death termed the hypersensitive response (HR) [ 6 ]. Breeding cultivars carrying R genes against pathogens is a common and powerful tool to control disease. However, the massive deployment of single R genes in the field exerts a strong selection pressure against avirulent pathogens that can become virulent through evolution of their AVR gene repertoire. Mechanisms leading to virulence include deletion, inactivation or down-regulation of the AVR gene, point mutations allowing recognition to be evaded while maintaining the virulence function of the AVR protein, or the acquisition of new effectors that suppress ETI [ 6 – 8 ]. Suppression of ETI by a fungal effector represents an efficient way to evade the selection pressure exerted by R genes in the field while maintaining the function of non-dispensable effectors. In some cases, the effector that suppresses ETI can itself be recognized by an R protein. A few examples of such strategies have been described in fungi [ 9 – 11 ], but the underlying mechanisms for the suppression of ETI by fungal effectors remain unexplained.

Wild type isolates Nz-T4 (a3a4a7), IBCN18 (a3A4A7) and v45.15 (a3a4a7), as well as Nz-T4, IBCN18 and v45.15 transformants carrying ECP11-1, and Nz-T4 transformants carrying both ECP11-1 and AvrLm4-7 were inoculated onto cotyledons of a cultivar carrying Rlm3 (15.22.4.1, A) or Rlm4 (Pixel, B). Pathogenicity was measured 13 days post-inoculation. Results are expressed as a mean scoring using the IMASCORE rating comprising six infection classes (IC), where IC1 to IC3 correspond to resistance, and IC4 to IC6 to susceptibility [ 50 ]. Error bars indicate the standard deviation of technical replicates. 19.4.24 (A3a4a7) and JN3 (a3A4A7) were used as controls of the AvrLm3/Rlm3, and AvrLm4-7/Rlm4 and Rlm7 interaction phenotypes, respectively.

As a starting point, a construct containing the ECP11-1 coding sequence and terminator under the control of the AvrLm4-7 promoter was introduced via Agrobacterium tumefaciens-mediated transformation into two isolates virulent towards Rlm3, Rlm4 and Rlm7: Nz-T4 and v45.15. The corresponding transformants (10 Nz-T4-ECP11-1 and 9 v45.15-ECP11-1) were inoculated onto B. napus cvs Pixel (Rlm4) and 15.22.4.1 (Rlm3; Fig 6 ). All transformants, as the wild type isolates, were virulent on Pixel. In contrast, and differing from their parental isolates, all transformants were avirulent on 15.22.4.1. We conclude that Ecp11-1, like AvrLm3, can be recognized by Rlm3. To confirm this result, we inoculated five transformants on two additional and unrelated oilseed rape cultivars carrying Rlm3 (Grizzly and Columbus) and confirmed a resistance phenotype on these cultivars triggered by Ecp11-1 ( S5 Fig ). Moreover, we also inoculated the transformants on the oilseed rape line 15.23.4.1, a sister line of 15.22.4.1 also issued from individual plants from cv. Rangi, carrying Rlm7 instead of Rlm3, and obtained a susceptibility phenotype ( S5 Fig ). These results support our conclusions that Ecp11-1 can be recognized by Rlm3.

We investigated whether the WR(F/L/V)(R/K) sequence motif, which is well conserved within the LARS structural family, could be involved in the ability of AvrLm4-7 to suppress recognition of AvrLm3 by Rlm3. Previous studies performed site-directed mutagenesis on AvrLm4-7 residues to investigate its ability to suppress the recognition of AvrLm3 by Rlm3 and to trigger Rlm7 and Rlm4-mediated immunity [ 22 , 24 ]. We extended these data with additional mutagenesis experiments ( Table 2 and S4 Fig ). Mutagenesis was performed on an allele of AvrLm4-7 conferring both Rlm7 and Rlm4-mediated recognition or only Rlm7-mediated recognition (G 120 R mutation). These mutations involved surface-exposed residues and are hence not expected to affect the global structure of AvrLm4-7 ( Fig 3D ). Mutations R 100 P and F 102 S led to a switch to virulence towards Rlm7 cultivar and abolished the ability of AvrLm4-7 to suppress Rlm3-mediated recognition of AvrLm3 when they were combined with a G 120 R mutation, suggesting that both R 100 or F 102 and G 120 are necessary to mask AvrLm3 recognition. In contrast, mutation S 112 R, located close to G 120 , was sufficient to escape Rlm4 and Rlm7-mediated recognition and abolish the ability of AvrLm4-7 to suppress Rlm3-mediated recognition of AvrLm3.

A sequence logo derived from the multiple alignment of the structural analogues highlighted conserved features of the LARS effectors ( Fig 5B ). Although the number of cysteines is variable between the different members, six cysteines can be aligned between the majority of the LARS members. The cysteines near the N- and C-termini form a disulfide bridge in all available structures, and this is likely the case for all of the proteins identified. The remaining aligned cysteines are not always involved in superimposable disulfide bridges in the three available structures. All structures have a putative disulfide bridge that connects α-helix 1 to β-strand 3. The third conserved cysteine pair establishes a different disulfide bridge, however, in AvrLm4-7 on the one hand and AvrLm5-9, Ecp11-1 and AvrLm3 on the other ( Fig 2A ). Apart from the cysteines, a WR(F/L/V)(R/K) sequence motif is very well conserved in all sequences ( Fig 2D ), positioned at the exit of the third β strand. The motif on one side crosses the disulfide bridge that connects the N- and C-termini, and on the other lies against the α-loop that precedes the second strand. Residues of this strand are less well conserved.

(A) Species distribution of number of structural analogues of AvrLm4-7, AvrLm5-9 and Ecp11-1 found with a low-stringency HMM search on a database encompassing 163 predicted proteomes of 116 fungal species. (B) Multiple sequence alignment of the 50 potential structural analogues and the sequences of the known structures (AvrLm4-7 = 4fprA / AvrLm5-9 = 0a59A / Ecp11-1 = 0cp1A). The secondary structures of the latter were calculated using the software STRIDE and were added at the bottom of the alignment (H = helix, G = 3–10 helix, E = β strand, B = β bridge, C = coil, T = turn) above the residue conservation measure, the local alignment quality and the consensus logo. The alignment was displayed using the software Jalview. In the displayed alignment, amino acids which do not align with any of the three sequences AvrLm4-7, AvrLm5-9 and Ecp11-1 have been removed. Cg, Colletotrichum gloeosporioides; Ch, Colletotrichum higginsianum; Co, Colletotrichum orbiculare; Cc, Corynespora cassiicola; Ff, Fulvia fulva; Lbb, Leptosphaeria biglobosa ‘brassicae’; Lbt, Leptosphaeria biglobosa ‘thlaspii’; Lmb, Leptosphaeria maculans ‘brassicae’; Lml, Leptosphaeria maculans ‘lepidii’; Mp, Macrophomina phaseolina; Pt, Pyrenophora teres; Ptr, Pyrenophora tritici-repentis; Sl, Stemphylium lycopersici.

To find out whether the LARS effector family has members in other fungi, a new HMM-based profile search was performed on an in-house database, composed of annotated proteomes of 163 fungal strains, corresponding to 116 species, mostly Dothideomycetes and Sordariomycetes with contrasting lifestyles (phytopathogens, entomopathogens, endophytes, saprophytes, mycoparasites, S3 Table ). The in-house database was iteratively searched with each effector structure file, using a cut-off E-value of 1 and a cut-off overlap of 50%. At each iteration, proteins longer than 160 amino acids were removed. This HMM-search identified, after three iterations, 34 potential structural analogues using AvrLm4-7, 32 with Ecp11-1 and two using AvrLm5-9. Interestingly, Ecp11-1 was found using AvrLm5-9 as a template, but not using AvrLm4-7. Combined with the analogues found in L. maculans ‘brassicae’, 49 non-redundant proteins were identified ( Fig 5 ). These potential structural analogues originate from 13 fungal species, with the majority from species closely related to L. maculans ‘brassicae’ (L. maculans ‘lepidii’, L. biglobosa ‘brassicae’ and L. biglobosa ‘thlaspii’), Dothideomycetes (Macrophomina phaseolina, Pyrenophora tritici-repentis, P. teres, Corynespora cassiicola, Stemphylium lycopersici and F. fulva), and a few Sordariomycetes (Colletotrichum orbiculare, C. gloeosporioides and C. higginsianum; Figs 5A and S2 , and S3 Table ).

The first pattern represented genes that were highly expressed as early as 5 days post-inoculation (DPI) with a peak at 7 DPI, and included the AVR genes of the family: Lmb_jn3_00001 (AvrLm3), Lmb_jn3_03262 (AvrLm4-7), Lmb_jn3_03263, Lmb_jn3_03815, Lmb_jn3_08343 (AvrLmS-Lep2), Lmb_jn3_10106 (AvrLm5-9) and Lmb_jn3_12986. The second pattern of genes was characterized by a low expression at 5 DPI and a plateau between 7 and 9 DPI: Lmb_jn3_01426, Lmb_jn3_01427 and Lmb_jn3_01428. The last pattern grouped genes whose expression peaked lower at 7 days post-inoculation: Lmb_jn3_08418, Lmb_jn3_08419 and Lmb_jn3_08421. Of interest, genes with the same expression profile are neighbors in the L. maculans genome. Finally, on V8 medium, all the genes were very lowly expressed, which indicates that they are overexpressed during the primary infection of oilseed rape by L. maculans.

The characteristics of LARS structural analogues identified in L. maculans using a HMM search are summarized in Table 1 . The genomic location of the genes encoding structural analogues was investigated using the latest assembly of the L. maculans genome [ 28 ]. The majority of these genes are located in genomic regions rich in remnants of transposable elements, with the exception of a group of three neighboring genes (Lmb_jn3_01426, Lmb_jn3_01427 and Lmb_jn3_01428), located in a gene-rich region, and a gene that had been previously described as a paralogue of AvrLm4-7 (Lmb_jn3_03263; [ 18 ]), which is located at the border of a gene-rich region. These genes are mostly located on different Super Contigs. However, several genes are neighbours. This is the case for (i) Lmb_jn3_01426, Lmb_jn3_01427 and Lmb_jn3_01428; (ii) Lmb_jn3_08418, Lmb_jn3_08419 and Lmb_jn3_08421; (iii) Lmb_jn3_03262 (AvrLm4-7) and Lmb_jn3_03263. The average size of these proteins is 140 amino acids, and they all have between six and 10 cysteines. For the whole family, apart from Lmb_jn3_08419, we were able to predict a signal peptide, suggesting that these proteins are secreted by the fungus.

We then wanted to confirm that the retrieved protein analogues have similar structures. We therefore constructed 3D structural models of the different analogues found by the HMM search using the AlphaFold2 program via the colab server ( https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb ; [ 27 ]). For all retrieved sequence analogues, AlphaFold2 proposed models that had the same fold as AvrLm4-7 (with the exception of Lmb_jn3_08343 and Lmb_jn3_12986 for which no reliable models could be proposed; S3 Fig ). Moreover, the cysteine bridges in all these models were superposable. We conclude that these sequences possess the same fold characterized by an antiparallel β-sheet and a characteristic set of disulfide bonds. This suggests they form a homologous family that we will name from this point forward as the LARS (for Leptosphaeria AviRulence and Suppressing) effector family.

The strong structural similarities between AvrLm4-7, AvrLm5-9, AvrLm3 and Ecp11-1 suggest that these four proteins belong to a fungal effector family, characterized by a four-stranded β-sheet, an α-helix and three conserved disulfide bridges ( Fig 2A and 2C ). To identify other structurally related family members, a hidden Markov model (HMM)-based profile search was performed on the predicted protein repertoire from L. maculans (v23.1.3 isolate). An iterative search was carried out with each effector structure, using a cut-off E-value of 1 and an overlap cut-off of 50%. A total of three iterations were performed. At each iteration, proteins longer than 160 amino acids were removed. Seven potential structural analogues were found for AvrLm4-7, six for Ecp11-1 and four for AvrLm5-9. The search with Ecp11-1 retrieved both AvrLm3 and AvrLm4-7, the remaining four were also found with the search using AvrLm4-7. The search with AvrLm5-9 retrieved AvrLm3 together with other candidates that were not found with AvrLm4-7 or Ecp11-1, including the AVR protein AvrLmS-Lep2 ( Table 1 ; [ 16 ]). A total of thirteen structural analogues were thus identified in L. maculans ( Table 1 ).

Interestingly, substitutions at I 34 and G 107 are always coupled and likely are responsible for the virulent phenotype towards Rlm3. They are positioned in the regions connecting the strands, and the folding brings them relatively close together in space, suggesting this site could be important for its function. Remarkably, the positions in the 3D structures of the polymorphic residues G 131 R (AvrLm3, G 107 in Fig 3 ) and G 120 R (AvrLm4-7; [ 24 ]) are very similar, suggesting the same protein regions could be involved in the virulence phenotype.

Sequencing of AvrLm3 in a collection of worldwide L. maculans isolates defined 22 different alleles, corresponding to 14 non-synonymous mutations leading to 11 isoforms of the protein [ 22 ]. We projected eight of the 11 polymorphic amino acids onto the AvrLm3 3D structure (the remaining three are located in the signal peptide and thus could not be plotted) ( Fig 3B ). Most of the AvrLm3 variants concern amino acid positions with surface-exposed side chains, and none of the variants affects the cysteines. Only one polymorphic substitution (L 78 to F) (L 54 according to protein numbering in Fig 2 ) is found in a residue involved in hydrophobic packing, but the mutation is conservative. These observations suggest that all alleles should lead to well-folded proteins. Six of the 11 AvrLm3 isoforms were only found in isolates avirulent towards Rlm3. These amino acids are scattered over the surface of the protein. Only two amino acid residues consistently differed between virulent and avirulent isoforms: I/L 58 H and G 131 R (I 34 and G 107 according to protein numbering in Fig 2 ). Of interest, amino acid residue I 58 is located in the same region of the structure in AvrLm3 as residue K 55 (responsible for the switch to virulence towards Rlm9) in the AvrLm5-9 structure. Similarly, residue G 131 in AvrLm3 is located in the same region as residue G/R 120 in AvrLm4-7 (responsible for the switch to virulence towards Rlm4 [ 18 ]; Fig 3D ; G 99 according to Fig 2 ).

We then set out to identify regions putatively required for recognition by cognate R proteins by mapping polymorphic residues onto their respective structures. We therefore exploited previous population studies that had reported polymorphisms in AvrLm5-9, AvrLm3 and Ecp11-1. Only three polymorphic residues of AvrLm5-9 were reported in L. maculans populations [ 14 , 21 ]. One event caused a switch to virulence towards both Rlm5 and Rlm9 (R 29 to stop), a point mutation at residue R 55 to T or K leads to virulence towards Rlm9 (K 36 according to the protein version and numbering in Fig 2 , since the AvrLm5-9 protein whose structure was determined conferred virulence towards Rlm9). A third polymorphism had no effect on the interaction with Rlm5 or Rlm9 (R 38 to L) (R 19 according to protein numbering in Fig 2 ). The stop mutation evidently leads to a truncated and non-functional protein. The two last polymorphic residues were mapped onto the AvrLm5-9 3D structure ( Fig 3A ). The mutation leading to virulence towards Rlm9 is situated in the middle of the long helix, and the mutation having no effect on the interaction with Rlm9 and Rlm5 is in a loop connecting β1 and β2. Neither mutation is expected to perturb the 3D structure.

A search in the Protein Data Bank (PDB) for structural analogues of Ecp11-1 and AvrLm5-9 using the DALI webserver returned AvrLm4-7 and the yeast elongation factor 1B. The latter protein (100 residues) indeed has the same β-sheet topology and similar connections, but shares no significant sequence identity with the effectors, has no disulfide bridges and the region of EF1B relevant for its function is not conserved in Ecp11-1 and AvrLm5-9. The similarity is therefore probably not biologically relevant.

The sequence similarities between AvrLm5-9, AvrLm4-7 and Ecp11-1/AvrLm3 are relatively weak (between 18 and 28%, Fig 2A and 2B ). Nevertheless, the 3D structures of all these effectors are strongly related: AvrLm5-9 and Ecp11-1/AvrLm3 share the same fold with AvrLm4-7 ( Fig 2B ), which displays suppressive interactions with AvrLm5-9 and AvrLm3. Both AvrLm4-7 (β4) and AvrLm5-9 (β2) have an irregular β-strand, but these are oriented and positioned in the same way as the corresponding strands in Ecp11-1. To fully appreciate the similarities between the three effectors, we superposed their structures using the DALI webserver ( http://ekhidna2.biocenter.helsinki.fi/dali/ ). Superposition of Ecp11-1 and AvrLm5-9 gives a Z-score of 8.9 and an RMSD value of 3.3 Å (106 residues aligned) and superposition of Ecp11-1 and AvrLm4-7 gives a Z-score of 5.6 and an RMSD value of 3.8 Å (106 residues aligned). The main differences between the three proteins are found in the regions connecting the strands ( Fig 2B and 2C ). The structures of the three effectors are stabilized by disulfide bridges, two of which are overlapping in the three proteins. All three effectors have a disulfide bridge that connects the N- and C-termini, and also at least one disulfide bridge that links the helical connection between β1-β2 and strand β3.

The presence of a metal, in this case Zn 2+ , was also mandatory for the crystallization of Ecp11-1. Speculating that the Ecp11-1 crystal would also bind a bivalent ion, we successfully solved its structure at 1.6 Å resolution by using the SAD-signal of Zn 2+ . The crystals contained one copy of Ecp11-1 in the asymmetric unit. The complete sequence could be positioned into the electron density. The analysis of the anomalous signal revealed the presence of two Zn-clusters in the structure that resemble those of the Ni-ions in AvrLm5-9. One Zn-cluster is bound near the N- and C-termini and involves histidines from the native protein and from the linker peptide and side chains from a crystal neighbor. Another Zn-cluster is found at the opposite end and is composed by residues emanating from neighboring Ecp11-1 molecules. Ecp11-1 forms an anti-parallel four-stranded β-sheet with (+2,-1,-2) topology (Figs 1B and 2C ). Strands β1 and β2 are connected by a peptide composed of a helical turn and a short helix surrounded by irregular peptides. The long connection between strands β3 and β4 contains a β-hairpin, a short helix and some irregular peptide stretches. Ecp11-1 contains five disulfide bridges: C 3 -C 137 (named SS1) that connects the N- and C-termini, C 22 -C 71 (named SS2) and C 26 -C 73 (named SS3) that link the β1-β2 connection to strand β3. The two remaining disulfide bridges (C 44 -C 96 and C 62 -C 65 , named SS4 and SS5) (according to protein numbering in Fig 2 ) are found in the irregular loop regions.

(A) Structure-based protein sequence alignment of AvrLm4-7, AvrLm5-9, AvrLm3 and Ecp11-1. The S-S bridges of Ecp11-1 are labelled green and numbered to show their connectivity. Secondary structure for each protein (β-sheets, α-helices, and β-turns are rendered as arrows, squiggles and TT letters respectively) is shown above the alignment. Identical residues are in red boxes and similar residues are in red. The conserved motif WR(F/L/V)(R/K) is labeled by black stars. The residues for whom mutations are associated with a switch to virulence are in green boxes. The figure was made with the ESPript server [ 66 ]. (B) Superposition of AvrLm4-7, AvrLm5-9 and AvrLm3. The variable connections between β3 and β4 are coloured in red (AvrLm3), blue (AvrLm4-7) and green (AvrLm5-9). The conserved or neighbouring S-S bridges are indicated as yellow, orange and cyan spheres (indicated as 1, 2 and 3, respectively, in panel A). (C) Representation of conserved topology, with the variable connexion between β3 and β4 coloured in red. (D) Superposition of AvrLm5-9, Ecp11-1 and AvrLm3 (all in grey), with their conserved WR(F/L/V)(R/K) motif at the exit of β3 represented by red sticks. A zoom on the motif is represented in the lower right corner.

After cleavage by the TEV protease and removal of thioredoxin, both AvrLm5-9 and Ecp11-1 provided good quality crystals. The structure of AvrLm5-9 was solved using iodine single wavelength anomalous diffusion (SAD) signal from a derivative crystal and then refined at 2.14 Å resolution using native data ( S1 Table ). The crystallization liquor of AvrLm5-9 contained 80 mM of Ni 2+ ions that proved mandatory for obtaining crystals. The structure revealed the presence of three Ni 2+ ions that are involved in crystal contacts: two are found at the N- and C-termini of AvrLm5-9 as a ligand with two histidines from the native protein, by one histidine from a linker peptide and one from a crystal neighbor. The third Ni 2+ ion is bound at the opposite end, and also interacts with histidines from two neighboring copies of AvrLm5-9 in the crystal. This suggests the bound Ni 2+ ions are important for crystal contacts and crystallization but unlikely have any functional relevance. The complete AvrLm5-9 sequence could be placed into the electron density, which also accounted for two residues from the linker peptide at the N-terminus. The structure of AvrLm5-9 consists of a central β-sheet made of three anti-parallel β-strands ( Fig 1A ). An elongated peptide (residues 54 to 64) runs anti-parallel to β-strand 2, but only establishes a few main-chain H-bonds, and therefore is not categorized as a β-strand. One face of the β-sheet is covered by the long connections between the stands. The connection between the β1 and β2 strands is a curved α-helix and the connection between the β3 and β4 strands contains a shorter helix surrounded by two irregular peptide loops. AvrLm5-9 has three disulfide bridges, which are C 3 -C 119 , C 22 -C 69 and C 26 -C 71 , named SS1, SS2 and SS3, respectively, according to protein numeration without the signal peptide ( Fig 2 ) or C 22 -C 138 , C 41 -C 88 and C 45 -C 90 according to protein numeration with the signal peptide. The SS1 bridge knits the N- and C-termini together, while the other two disulfide bridges fix the long helix onto the β-sheet.

A small secreted protein with 37% amino acid sequence identity with AvrLm3 was identified from F. fulva [ 26 ]. This protein, named Ecp11-1 (Extracellular protein 11–1), was found in apoplastic washing fluid samples harvested from compatible F. fulva–Solanum lycopersicum (tomato) interactions. Curiously, Ecp11-1 triggers an HR in multiple wild accessions of tomato. It is therefore likely that Ecp11-1 is an AVR effector recognized by a corresponding R protein (tentatively named Cf-Ecp11-1) in wild accessions of tomato [ 26 ]. We decided to produce Ecp11-1 using the same P. pastoris-based strategy as described for AvrLm5-9 and AvrLm3. The yields of Ecp11-1 production were sufficient to start structural studies ( S1 Fig , i.e. about 5 mg of purified Ecp11-1 per liter of cell culture).

To explore the putative molecular relationships between AvrLm4-7, AvrLm5-9 and AvrLm3, we set out to determine the 3D structures of the AvrLm3 and AvrLm5-9 effectors. AvrLm5-9 and AvrLm3 are rich in cysteines and therefore are difficult to express in soluble form in Escherichia coli. For the recombinant production of AvrLm5-9 and AvrLm3, we therefore chose the well-established Pichia pastoris eukaryotic expression system. The genes coding for the AvrLm5-9 and AvrLm3 proteins without their secretion signal peptides were cloned into expression vectors as fusion proteins with a purification His-tag and with thioredoxin. A TEV proteolytic cleavage site was inserted between thioredoxin and the effectors. The AvrLm5-9 fusion protein was well expressed and purified to homogeneity ( S1 Fig ), but the yields of the AvrLm3 fusion protein were insufficient (i.e. about 50 mg of pure AvrLm5-9 per liter of cell culture against less than 1 mg of purified AvrLm3).

Discussion

In this study, we determined the crystal structure of L. maculans AvrLm5-9 and F. fulva Ecp11-1, and obtained a good quality model for AvrLm3 built via the crystal structure of Ecp11-1. Despite their poor sequence similarity, these three effectors are structural analogues of AvrLm4-7. All have a four-stranded β-sheet and helical connections with the same topology. The main differences reside in the conformations of the connections between the strands. Six cysteines involved in disulfide bridges are shared by the three effectors. One disulfide bridge ties together the N- and C-terminal regions, and two others connect the main helical region to the β-sheet. Structure-based pattern searches identified a large number of LARS effector candidates displaying sequence diversity, but likely sharing the same fold. Sequence alignment and 3D model superposition obtained using AlphaFold2 of these candidates shows the strong conservation of six cysteines, which are involved in the aforementioned structure stabilizing disulfide bridges. All of the retrieved sequence analogues are likely compatible with the structures, as confirmed by the structure prediction server I-TASSER.

The alignment of the putative analogues highlights a conserved sequence patch, WR(F/L/V)(R/K), with (F/L/V) being a hydrophobic/aromatic residue. These residues are situated at the end of the third β-strand, close to the N and C termini. The tryptophan and arginine are solvent exposed and could provide an interaction surface with plant targets. The hydrophobic (F/L/V) residue is involved in hydrophobic packing of this patch against the β-sheet. Interestingly, site-directed mutagenesis of R100 or F102 residues resulted in the loss of Rlm7-mediated recognition and abolished the ability of AvrLm4-7 to mask the recognition of AvrLm3 by Rlm3. The latter, however, only occurred in combination with the G120R mutation which allowed the effector to escape Rlm4-mediated recognition. In contrast to these results, a mutation at residue S112 allowed AvrLm4-7 to escape Rlm4 and Rlm7-mediated recognition but also abolished the ability of this effector to mask AvrLm3 from Rlm3-mediated recognition. The polymorphic residues identified in AvrLm5-9 and AvrLm3 from L. maculans populations and on Ecp11-1 from F. fulva isolates are mainly located on the loop regions of the proteins, with the only exceptions being a few amino acid changes (putatively) involved in the switch to virulence towards cultivars with Rlm3 or Rlm9 R genes. Remarkably, the positions in the 3D structures of the polymorphic residues G131R in AvrLm3 and G120R in AvrLm4-7 are very similar, as are I/L58H in AvrLm3 and R55K in AvrLm5-9, suggesting the same protein regions could be involved in the virulence phenotypes.

Structure-informed pattern searches specifically identified LARS-effectors in phytopathogenic ascomycetes from the Dothideomycetes and Sordariomycetes classes. One or two LARS-effector(s) per species were detected in the Sordariomycetes Colletotrichum spp. Two structural analogues were identified in the Dothideomycetes closely related to L. maculans ‘brassicae’, specifically P. tritici-repentis and P. teres, while between three and four structural analogues were identified in the species from the complex comprising L. maculans ‘brassicae’, suggesting a recent expansion of the LARS family in the species complex. In L. maculans ‘brassicae’, thirteen LARS effectors could be detected and their expression during the primary biotrophic stages of oilseed rape cotyledon infection suggests they are bona fide effectors. They represent 14% of the candidate effectors specifically overexpressed during the biotrophic stages of oilseed rape infection (nine LARS effectors among the 63 effector genes in Cluster 2 ‘biotrophy’ defined by Gay et al. [29]). The LARS family also comprises four out of the nine cloned AVR genes from L. maculans. Most of the L. maculans ‘brassicae’ LARS effectors are located in TE-rich regions (9/13) and eight are grouped in three genomic regions as neighbor genes, suggesting their expansion could be partly due to local duplications and that their location in TE-rich compartments could have led to their rapid diversification [30]. Expansions of the LARS family, comprising between five to eight structural analogues, were also detected in two other Dothideomycetes, M. phaseolina and C. cassiicola. We conclude that LARS effectors probably have a common evolutionary origin and that their expansion in some Dothideomycetes results from duplications, and, at least in the L. maculans ‘brassicae’ genome, diversification in TE-rich compartments.

Four other structural families of effectors were reported in fungi: the RALPH effectors identified in Blumeria graminis, the MAX effectors identified in Magnaporthe oryzae, the FOLD effectors identified in F. oxysporum and the ToxA-like family first identified in P. tritici-repentis (RALPH for RNAse-Like Proteins Associated with Haustoria, [31]; MAX for Magnaporthe Avrs and ToxB like, [32]; FOLD for Fusarium oxysporum f. sp. lycopersici dual-domain, [33]; [34]). The RALPH family represents about 25% of the B. graminis predicted effectors and three out of the four AVR effectors identified to date, and most of them are highly expressed during plant infection [10,31,35,36]. Pedersen et al. [31] hypothesized that RALPH effectors originated from an ancestral gene, encoding a secreted ribonuclease, duplicated by TE-driven processes and recently diversified within the grass and cereal powdery mildew lineage. The same way, the MAX family represents between 5 to 10% of the M. oryzae effectors and 25% of the cloned AVR effectors, and most of them are expressed during early biotrophic stages of rice infection. Such an expansion has also been observed in Venturia inaequalis, where a family of 75 in planta-upregulated MAX effectors has been identified [37]. De Guillen et al. [32] hypothesized that, in the case of Magnaporthe, the expansion of the MAX family occurred in a common ancestor of M. oryzae and M. grisea. Recently, Yu et al. [33] described a new structural family identified in F. oxysporum f. sp. lycopersici for four SIX (Secreted In Xylem) effectors, SIX4 (AVR1), SIX1 (AVR3), SIX6 and SIX13. Three V. inaequalis effectors were also recently predicted to belong to the FOLD family [37]. Finally, ToxA, which was first described in P. tritici-repentis, was found to belong to a structural family (ToxA-like) with members in M. lini, F. oxysporum f.sp. lycopersici and V. inaequalis [33,34,37–39]. The scenario observed for the LARS, RALPH, MAX, FOLD and ToxA-like examples suggests that a wide variety of effectors, without any apparent sequence relationship, could in fact constitute a limited set of structurally conserved effector families and that they have expanded in some fungal lineages or even in several fungal classes.

AvrLm4-7 suppresses Rlm3- and Rlm9-mediated disease resistance. Other cases of effectors with a suppressive function have been described in fungi. In Fusarium oxysporum f. sp. lycopersici, for instance, the Avr1 effector suppresses plant immunity mediated by the I-2 and I-3 R proteins of tomato, which recognize the Avr2 and Avr3 effectors, respectively [11]. The structure of Avr2 has recently been determined [38], but is unrelated to the structures of AvrLm4-7, AvrLm5-9 or AvrLm3. Interestingly, the Avr1 and Avr3 structures were recently solved by crystallography and shown to belong to the FOLD family [33]. It has been suggested that Avr1 may suppress I-3-mediated immunity by preventing Avr3 recognition through competitive inhibition [33]. In another example, the necrotrophic fungus P. tritici-repentis, the Host Selective Toxin (HST) ToxA suppresses the activity of other HSTs [40], while in B. graminis, a suppressor of avirulence (SvrPm3) has been identified that acts on the interaction between the AVR gene AvrPm3 and the barley R gene Pm3 [10]. In both of these examples, like the example of the FOLD effectors described above, the mechanisms underlying suppressive function have yet to be determined.

In non-fungal models, several mechanisms explaining suppressive interactions have been highlighted. (i) The AVR effector can act downstream of another AVR recognition by an R protein to suppress HR induction: in Xanthomonas campestris pv. vesicatoria, AvrBsT interacts in the plant cell with SnRK1, an SNF1-related kinase, to inhibit the HR induced by AvrBs1 recognition [41]. (ii) Effectors displaying suppressive interaction can share a common plant target, but differ in their actions on that target: in Pseudomonas syringae, the effector proteins AvrRpm1, AvrRpt2 and AvrB target the Arabidopsis thaliana protein RIN4, a key regulator of plant immunity [42,43]. While AvrB and AvrRpm1 trigger Rpm1-mediated recognition through phosphorylation of RIN4, AvrRpt2 triggers plant immunity through cleavage of RIN4, thus preventing recognition of AvrB and AvrRpm1 by Rpm1. (iii) Suppressive effectors can directly act on the R proteins: in Phytophthora infestans, the IPI-O4 effector suppresses the HR triggered by recognition of IPI-O1 by the potato RB R protein. IPI-O4 interacts with the coiled-coil domain of RB which is also the domain targeted by IPI-O1 [44]. Since AvrLm4-7, AvrLm3 and AvrLm5-9 share the same structural fold, we hypothesize that they could target the same plant components or cellular processes and / or be recognized by the same R proteins.

So far, we do not have any information on the plant components targeted by AvrLm3, AvrLm5-9 and AvrLm4-7, since we did not identify any relevant or common plant target by performing yeast two-hybrid (Y2H) assay screening on a cDNA library of oilseed rape infected by L. maculans. However such a target could be guarded by the R proteins or, in the case of a direct interaction with the R proteins, could be Rlm9 or Rlm3 themselves. Rlm9 was cloned and found to encode a wall-associated kinase-like (WAKL) protein, a newly described class of Receptor-Like Kinase (RLK) R protein [45]. Using a Y2H assay, no direct interaction between the extracellular region of Rlm9 and AvrLm5-9 could be detected. However, a direct interaction between AvrLm5-9 and Rlm9 cannot be excluded since Y2H is not an optimal technique to test interaction with a membrane protein. Notably, Haddadi et al. [46] recently cloned Rlm4 and Rlm7 and found they corresponded to alleles of Rlm9, with the three encoded proteins only differing by a few amino acid residues in the extracellular receptor domain of the WAKL. Larkan et al. [45] had also previously identified at the Rlm9 locus, in another resistant accession of oilseed rape, a WALK gene that could potentially correspond to Rlm3, and being allelic to Rlm9. Another example of an allelic R protein that is able to recognize sequence-unrelated AVR effectors with a predicted common fold was recently reported in barley [47], the specificity of recognition being conferred by amino-acid modifications in the LRR domain of the MLA R protein. However, it is currently unknown whether MLA directly interacts with B. graminis effectors. Based on the zig-zag model [6], we hypothesize that Rlm4 and Rlm7 evolved from Rlm3 or Rlm9 in response to the suppressive effect of AvrLm4-7 on AvrLm3 and AvrLm5-9 recognition. We propose a model in which Rlm3 and Rlm9 directly recognize the complex between AvrLm3 (and Ecp11-1) or AvrLm9 and their plant target (or AvrLm3 and AvrLm9 themselves after they have bound to their plant target). AvrLm4-7 would have a higher affinity for the same host target, thereby preventing interaction with AvrLm3 or AvrLm5-9, and thus masking their presence to Rlm3 and Rlm9. Although AvrLm4-7 binds the same host virulence target, we hypothesize that it does not possess the protein region recognized by Rlm3 and Rlm9. Instead, upon binding the plant target, AvrLm4-7 presents a protein region that is recognized by Rlm4 and Rlm7.

We have identified a large structural family of effectors that, in L. maculans, are expressed during the early stages of infection and are potentially targeted by R proteins. This structural information on effectors could be used to improve the management and durability of R genes in the field. Indeed, among the nine AVR genes identified to date in L. maculans, four belong to the LARS family. The corresponding R genes are, at least in part, present in commercial varieties currently used in the fields (Rlm7, Rlm3, Rlm4, Rlm9, RlmS). We hypothesize that the presence of R genes targeting members of the LARS family potentially exerts a selection pressure on the other members of the family, and that an efficient strategy to improve durability of R genes would consist in alternating or pyramiding R genes corresponding to different structural classes of effectors. We have also determined that Ecp11-1, a homologue of AvrLm3 and AVR effector candidate from F. fulva, is able to trigger Rlm3-mediated resistance in oilseed rape. This finding significantly alters our understanding about the degree of host-microbe specificity as developed by Flor in the 1940s [48], in that this is one of the few examples of cross-species effector recognition, as previously mentioned by Stergiopoulos et al. [49] for the recognition of the Avr4 effector from F. fulva and of its orthologue in Pseudocercospora (Mycosphaerella) fijiensis by the Cf-4 R protein of tomato. A next step will be to determine whether other homologues identified in Dothideomycetes and Sordariomycetes can also trigger recognition by R proteins of oilseed rape and, in the longer term, to evaluate the possible use of broad-spectrum resistances for multi-pathogen management of diseases.

[END]
---
[1] Url: https://journals.plos.org/plospathogens/article?id=10.1371/journal.ppat.1010664

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.