Unconventionally secreted effectors of two filamentous pathogens target plant salicylate biosynthesis

Plant diseases caused by fungi and oomycetes pose an increasing threat to food security and ecosystem health worldwide. These filamentous pathogens, while taxonomically distinct, modulate host defense responses by secreting effectors, which are typically identified based on the presence of signal peptides. Here we show that Phytophthora sojae and Verticillium dahliae secrete isochorismatases (PsIsc1 and VdIsc1, respectively) that are required for full pathogenesis. PsIsc1 and VdIsc1 can suppress salicylate-mediated innate immunity in planta and hydrolyse isochorismate in vitro. A conserved triad of catalytic residues is essential for both functions. Thus, the two proteins are isochorismatase effectors that disrupt the plant salicylate metabolism pathway by suppressing its precursor. Furthermore, these proteins lack signal peptides, but exhibit characteristics that lead to unconventional secretion. Therefore, this secretion pathway is a novel mechanism for delivering effectors and might play an important role in host–pathogen interactions.

P hytophthora sojae is an oomycete pathogen from soybean 1 , whereas Verticillium dahliae is a vascular wilt fungus that causes diseases in 4200 plant species 2 . Although they are evolutionarily distant eukaryotes, both are filamentous pathogens that grow or infect their hosts via thread-like hyphae. These filamentous plant pathogens secrete a complex repertoire of effector proteins to manipulate host innate immunity 3,4 . The effectors can either contribute to pathogen virulence by overcoming plant defense (termed effector-triggered susceptibility), or be directly or indirectly recognized by host surveillance systems, leading to effector-triggered immunity 5 . Typically, the secreted effectors are characterized by short N-terminal signal peptides that direct them to the secretory pathway 3,6 . P. sojae produces hundreds of apoplastic effectors and putative cell-entering effectors 1,3,6 , whereas V. dahliae secretes more than 700 putative effectors, including many small cysteinerich proteins that are thought to function in the host apoplast to promote disease 2 . For example, the Verticillium effector Ave1 contributes to host fungal virulence in the absence of its corresponding R protein (Ve1) 7 . Nevertheless, the virulence targets and mode of action of these effectors remain largely unknown.
Salicylate (SA) is a central signaling molecule during plant innate immune responses. It is usually induced by infection with biotrophic or hemibiotrophic pathogens, and serves as an important component of PAMP-triggered immunity (PTI), effector-triggered immunity and systemic acquired resistance 8,9 . Plants synthesize SA through two distinct enzymatic pathways. One is from phenylalanine via cinnamate and the other is from chorismate via isochorismate 8,10 . The second pathway is also shared by some bacteria. For example, Pseudomonas aeruginosa produces SA via two distinct enzymes: isochorismate synthase (ICS) and isochorismate pyruvate lyase 11 . ICS homologues have been identified in a wide variety of plant species and genetic analyses revealed that the ICS-dependent SA biosynthetic pathway predominates under (a)biotic stresses in Arabidopsis, although IPL genes have not yet been identified in plants 8,10 .
Isochorismate is a precursor of both SA and many other distinct derivatives in plants, fungi and bacteria 10 . For example, in bacteria, isochorismate can be converted into 2,3-dihydro-2,3dihydroxybenzoate (DDHB) and pyruvate by isochorismatase (ISC, EC 3.3.2.1). Escherichia coli EntB and Vibrio cholerae VibB contain a conserved amino-terminal ISC domain and carboxy-terminal aryl carrier protein domain. They play roles in the production of enterobactin and vibriobactin, respectively 12,13 . Pseudomonas aeruginosa PhzD is required for phenazine biosynthesis; it is 46% identical to EntB, but it lacks its C-terminal domains. PhzD also catalyses the hydrolysis of isochorismate 14 . Isochorismatase-like hydrolases have been identified from the pathogenic protozoans Leishmania donovani, L. major and Trypanosoma cruzi 15 , several phytopathogenic fungi 16 and the rice root nematode Hirschmanniella oryzae 17 . However, their functions in these pathogens remain unknown.
Here we identified two genes from P. sojae (PsIsc1) and V. dahliae (VdIsc1) whose putative encoded proteins share high similarity with known ISCs and exhibit ISC activity. We demonstrated that both genes were required for virulence and that they function by suppressing SA accumulation in their hosts. The genes encode unconventionally secreted proteins and PsIsc1 could mediate the translocation of the C terminus of the P. sojae effector Avr1b into host cells. The in planta expression of these genes reduced SA levels. We propose that these two unconventionally secreted effectors hydrolyse host isochorismate, thereby modulating the SA metabolism pathway and suppressing host immunity.

Results
PsIsc1 and VdIsc1 are putative ISCs. As ISC can hydrolyse isochorismate (the direct precursor of SA), we hypothesized that some phytopathogens might exploit ISCs as virulence effectors to suppress host SA biosynthesis. Therefore, we searched for P. sojae and V. dahliae ISCs using reciprocal BLAST searches with 40 known ISCs from the PDB database (Supplementary Table 1). Of 17 candidates, three each from P. sojae and V. dahliae were identified as putative ISCs based on HHpred searches 18 (Supplementary Table 2). Of these 6 candidate genes, one from each species (PsIsc1 and VdIsc1) was upregulated during the infection stages ( Supplementary Fig. 1). The predicted PsIsc1 and VdIsc1 proteins shared high primary and secondary structural similarities with the reported ISCs ( Supplementary Fig. 2). Thus, they are candidate virulence proteins in these two filamentous pathogens.
PsIsc1 is a virulence factor in P. sojae. To determine the role of PsIsc1 in P. sojae virulence, we silenced or overexpressed the gene using polyethylene glycol-mediated protoplast transformation 19 . The PsIsc1 transcripts were reduced to 38% and 18% in two silenced transgenic lines (ST6 and ST22, respectively) and increased by 90-and 376-fold in two overexpression transgenic lines (OT3 and OT20, respectively) ( Fig. 1a). Western blotting confirmed that PsIsc1 was successfully expressed at the expected size in OT3 and OT20 (Fig. 1b). All transformants showed normal filamentous growth. Compared with the wild-type (WT) strain and a transgenic line (T10) with a normal PsIsc1 expression level, the two silenced lines (ST6 and ST22) showed a visible reduction in virulence on soybean seedlings, whereas the two overexpressed lines (OT3 and OT20) were more virulent than the controls (Fig. 1c,d).
To further characterize PsIsc1-mediated virulence, we measured the SA levels in soybean seedlings infected with the P. sojae transformants. PsIsc1 silencing increased the concentrations of both free SA and its glucoside (SAG) by B3-fold. In contrast, PsIsc1 overexpression significantly decreased the concentrations of SA and SAG (Fig. 1e). The transcript levels of the SA-inducible PATHOGENESIS-RELATED 1 (PR1) gene were significantly increased or reduced accordingly at sites inoculated with the transformants compared with WT and T10 (Fig. 1f). Collectively, our results suggest that PsIsc1 functions as a virulence factor in P. sojae by suppressing host SA accumulation.
VdIsc1 is required for full virulence in V. dahliae. To explore the role of VdIsc1 in V. dahliae virulence, deletions were generated in V. dahliae strain V991 to produce two mutants: D1-12 (V991DVdIsc1-12) and D1-18 (V991DVdIsc1-18). Two mutants exhibited normal growth phenotypes ( Supplementary Fig. 4). Based on pathogenicity tests in the cotton cultivar Lumian 21, the two VdIsc1 deletion mutants produced significantly reduced disease symptoms. To confirm that the mutant phenotypes resulted from the deletion of VdIsc1, a single copy of the gene was introduced into D1-12. The resulting complemented strain, D1-12VdIsc1, was fully pathogenic and induced disease at rates similar to the WT strain. We also explored the ability of PsIsc1 to complement the D1-12 deletion. PsIsc1 could functionally replace VdIsc1 ( Fig. 2a and Supplementary Fig. 4), suggesting that the two genes play similar roles in virulence.
In cotton plants inoculated with the two deletion mutants, the free SA and SAG levels were increased by 100-150% (Fig. 2b) and there was a corresponding B2-fold increase in the level of cotton PR1 mRNA (Fig. 2c). The concentration of SA, SAG and PR1 transcript levels after infection with the VdIsc1-and PsIsc1complemented transformants were comparable with the WT (Fig. 2b,c). Thus, both genes can act as virulence factors in V. dahliae by reducing host SA levels. Both PsIsc1 and VdIsc1 contain a triad of catalytic residues (Asp-Lys-Cys; D-K-C), which form the activate sites and are responsible for the features of many known ISCs 15,20 (Supplementary Fig. 2). We generated substitution mutations of the three key amino acids in both genes (PsIsc1 A3 and VdIsch1 A3 ) and found that neither mutant could complement the V. dahliae DVdIsc1 mutant based on virulence tests (Fig. 2a), suggesting the virulence contributions of these effectors are probably linked to their enzymatic activities.
To elucidate the biological significance of the elevated SA levels during infection with the deletion mutants, we estimated the effect of SA on cotton resistance. The results revealed that pretreatment with SA reduced virulence, comparable to D1-12 infections ( Supplementary Fig. 5). This suggests that SA enhances the resistance of cotton toward V. dahliae. Furthermore, we analysed the virulence contributions of VdIsc1 using the Arabidopsis thaliana WT and NahG transformant plants. SA levels are extremely low in Arabidopsis plants that expresses the Pseudomonas putida NahG gene, because it encodes salicylate hydroxylase, an enzyme that converts SA to catechol 21 . As shown in Fig. 2d, the NahG transformant was significantly more susceptible to V. dahliae than A. thaliana WT, supporting the hypothesis that SA is required for plant resistance to V. dahliae. D1-12 or D1-12VdIsc A3 , but not D1-12VdIsc1, exhibited significantly reduced virulence in A. thaliana WT. In contrast, the NahG transformant showed comparable susceptibility to V. dahliae WT, D1-12, D1-12VdIsc1 and D1-12VdIsc A3 . Taken together, these results suggest that VdIsc1 suppress plant SA-mediated immunity, which could be a direct consequence of its inability to interfere with SA biosynthesis in the host plants.
PsIsc1 and VdIsc1 are unconventionally secreted proteins. To identify the site of action of VdIsc1 and PsIsc1 during interactions, we analysed whether they were secreted. Neither protein was predicted to contain an N-terminal signal peptide that could direct the protein to the conventional secretory pathway (Supplementary Table 2 and Supplementary Fig. 2). However, both were predicted to be non-classically secreted based on SecretomeP 2.0 (ref. 22), with NN scores (neural network output scores) of 0.66 and 0.81, respectively (Supplementary Table 2), which is above the threshold of 0.5. To test the predictions experimentally, we used western blotting to detect secreted proteins in culture supernatants of the P. sojae transgenic line OT20 and the V. dahliae complemented line D1-12VdIsc1, which produced His-tagged PsIsc1 and haemagglutinin (HA)-tagged  ARTICLE VdIsc1, respectively. PsIsc1-His and VdIsc1-HA could be detected in culture supernatants, as well as in hyphae of the respective transformants. However, the control cytoplasmic protein, a-actin, was only found in the hyphae of both cases (Fig. 3). Thus, PsIsc1 and VdIsc1, although they lack classical N-terminal signal peptides, are secreted proteins.

PsIsc1 is translocated into host cells during interactions.
To assess the localization of PsIsc1 during infection, we generated the P. sojae transformants producing the coding region of monomeric red fluorescent protein (mRFP) fused to the C terminus of PsIsc1 or the P. sojae RxLR effector Avr1b 19 . PsIsc1-mRFP fusion protein was validated in both mycelia and culture supernatants, and was the same as the Avr1b-mRFP. In contrast, soluble protein green fluorescent protein (GFP) was detected in mycelia, but not in the culture supernatants ( Supplementary Fig. 6). These results further indicate that PsIsc1 and Avr1b are secreted proteins. Avr1b-mRFP was predominantly localized in haustoria and the extrahaustorial matrix during soybean infection. Fluorescence was not detected within infected cells, presumably due to the dilution of Avr1b-mRFP within the plant cells ( Fig. 4a), as observed with P. infestans Avr3a 23 . PsIsc1-mRFP also showed strong localization in haustoria and the extrahaustorial matrix, but was not detected in plant cells. The unfused control (mRFP) did not show haustorial localization (Fig. 4a). Therefore, PsIsc1 is distributed in a similar manner as the host cytoplasmic effector Avr1b, and both proteins might be secreted from haustoria during interactions between P. sojae and soybean.
To determine whether PsIsc1 enters host cells following secretion, we examined whether it could deliver Avr1b into host soybean cells. Oomycete RxLR effectors such as the P. infestans effector Avr3a 23 and P. sojae Avr1b 19,24 depend on their N-terminal RxLR-dEER domain for host cell delivery. We demonstrated previously that the transformants expressing WT Avr1b lost the ability to infect soybean plants carrying Rps1b but retained the ability to infect plants lacking Rps1b, and that the N-terminal RxLR-dEER domain was required its translocation into host cells 19 . We deleted the N-terminal region from Avr1b (producing Avr1bCt) and replaced it with PsIsc1 (producing PsIsc1-Avr1bCt), and then generated two P. sojae transformants that expressed the fused PsIsc1-Avr1bCt (Fig. 4b). Production of the expected proteins was confirmed using western blotting (Fig. 4c). Transformants expressing the fused Avr1b-mRFP were consistently avirulent on soybean cultivars carrying Rps1b (Fig. 4d), suggesting that the recognition of Avr1b inside soybean cells was not affected by the fusion with mRFP. The inoculation   of soybean hypocotyls showed that the two transformants expressing PsIsc1-Avr1bCt were unable to infect the soybean cultivar HARO13 containing the Rps1b resistance gene, but could still infect Williams, which lacks Rps1b. In contrast, the recipient WT strain (P6497) and the transgenic lines expressing PsIsc1 (OT20) or Avr1bCt-mRFP could kill both cultivars (Fig. 4d). Similar results were observed using a leaf inoculation assay ( Supplementary Fig. 7). Taken together, these data suggest that PsIsc1 could functionally replace the secretory leader and host cell entry domain of Avr1b, and that PsIsc1 might be translocated inside host cells during infection.
PsIsc1 and VdIsc1 suppress SA-mediated defense in planta. To determine whether PsIsch1 and VdIsch1 function within plant cells to compromise immunity, we expressed these two genes (as GFP fusions) transiently in Nicotiana benthamiana tissue using Agrobacterium infiltration. Western blotting with anti-GFP antibodies confirmed the expression of proteins of the expected length (Fig. 5a). We also assessed the distribution of PsIsc1 and VdIsc1 in plant cells. Neither of the two proteins were co-localized with the mCherry-tagged plastid marker pt-rk CD3-999 (ref. 25), but were instead localized to the cytoplasm and nucleus of transformed N. benthamiana cells, similar to native GFP alone ( Supplementary Fig. 6). We then inoculated leaves with a compatible pathogen, P. capsici. Compared with leaves expressing GFP alone, the diameters of the lesions caused by P. capsici were significantly increased following expression of PsIsc1 and VdIsc1 (Fig. 5b,c). The levels of SA (Fig. 5d) and PR1 mRNA (Fig. 5e) in the inoculated sites were reduced when the two genes were expressed, indicating that PsIsc1 and VdIsc1 suppressed SA-mediated Phytophthora resistance in N. benthamiana. When mutations were introduced into the catalytic triads of PsIsc1 and VdIsc1, they no longer suppressed resistance to P. capsici or reduced SA and PR1 transcript levels when transiently expressed in N. benthamiana (Fig. 5).
PsIsc1 and VdIsc1 have ISC activity. To examine whether PsIsc1 and VdIsc1 could catalyse hydrolysis of isochorismate into DDHB, we purified targeted proteins from N. benthamiana leaves (Fig. 5a) and used E. coli EntB and V. cholerae VibB that were expressed in E. coli as positive controls. The enzyme activity was determined as described previously by measuring the increase in absorbance at 340 nm caused by NAD þ reduction 13 . In plantaexpressed PsIsc1 and VdIsc1, as well as two positive controls, exhibited high enzyme activity. Mutations in the catalytic triads of PsIsc1 and VdIsc1 resulted in an almost complete loss of enzyme activity. No activity was detected when GFP (a negative control) was purified and analysed using the same procedures (Fig. 5f). We also measured the relative levels of DDHB in plant tissues expressing the above genes. Significantly increased DDHB concentrations were detected in total extracts of the leaves expressing PsIsc1 or VdIsc1, whereas DDHB accumulated at SP Avr1bCt Avr1b Avr1b-mRFP  relatively lower levels in leaves expressing GFP, PsIsc1 A3 or VdIsch1 A3 (Fig. 5g). Therefore, both proteins have ISC activity.
N termini of PsIsc1 and VdIsc1 are required in V. dahilae. PsIsc1 and VdIsch1 contain no motifs that are similar to RxLR 19,23 . We constructed mutants of both proteins lacking the N-terminal regions (PsIsch1D Nt and VdIsch1D Nt ) ( Supplementary Fig. 2). In these mutants, the NN scores calculated using the SecretomeP 2.0 server were below the normal threshold for secreted proteins 22 (Supplementary Fig. 2). Neither of the deletion mutants complemented the phenotypes of VdIsc1-deletion mutants in cotton seedlings (Fig. 6a), suggesting that both proteins require their putative unconventional secretion signals to function in V. dahliae. To confirm that loss of the N-terminal regions did not compromise ISC activity, we transiently expressed PsIsc1D Nt and VdIsc1D Nt in N. benthamiana leaves. The deletion mutants conferred similar phenotypes as their WT in terms of the levels of resistance (Fig. 6b), the SA/SAG content and PR1 transcripts (Fig. 6c). Therefore, the N-terminal region is probably required for the secretion of these proteins from pathogens, but not for their roles in plant cells.

Discussion
Little is known about how filamentous plant pathogens interfere with host innate immunity for successful colonization and infection. As SA-mediated signalling pathways are important for plant immune responses and isochorismate is a precursor of SA, it has been inferred that phytopathogens could produce ISCs to reduce SA accumulation in response to pathogen attack and thus inhibit plant defense responses 16,17,26 . In the current study, we have demonstrated experimentally that P. sojae and V. dahliae encode ISC effectors (PsIsc1 and VdIsc1) to disrupt the SA metabolism pathway. Notably, we provide evidence that the ISCs are probably secreted from the two filamentous pathogens through unconventional pathways, and that PsIsc1 can enter into host cells during infection. It was reported that ISC family motifs were found in the secretome of some phytopathogenic filamentous ascomycete fungi, but not in non-phytopathogenic species 16 . Isochorismatase hydrolase was also identified specifically in a highly aggressive isolate of V. dahliae using proteomics-based analyses 26 . We identified putative ISC genes in P. sojae and V. dahliae (PsIsc1 and VdIsc1, respectively) based on primary and secondary structural similarities with known ISCs and demonstrated that both genes were highly expressed during infection stages. The overexpression or silencing of PsIsc1 in P. sojae led to increased and reduced virulence, respectively; SA levels in infected host tissues were altered accordingly. Similar results were observed when VdIsc1 was deleted in V. dahliae. Interestingly, the PsIsc1 gene could functionally replace VdIsc1 in V. dahliae. The transient overexpression of these two genes in N. benthamiana tissues suppressed SA levels and SA-mediated defense; consistent with this, DDHB levels were increased. Furthermore, we demonstrated that the conserved ISC catalytic An identical quantity of the tested proteins was purified from plant tissues. VibA generates NADH from DDHB and NAD þ when active ISC was present (producing DDHB from isochorismate); the absorbance of NADH at A340 nm at the indicated time points (seconds) was recorded. (g) Relative DDHB concentrations in plant tissues expressing the indicated genes. The relative DDHB concentration of each tested sample was calculated based on A340 absorbance and the DDHB concentration in GFP-expressing leaves was normalized to 1.0. In a, c-e and g, from left to right: GFP, VdIsc1, PsIsc1, VdIsc1 A3 and PsIsc1 A3 . **Po0.01 (Dunnett's test).
triads in PsIsc1 and VdIsc1 were required for ISC activity and the suppression of plant immunity. Thus, we conclude that these two evolutionary distinct pathogens both generate ISCs to promote infection by destroying a precursor (isochorismate) of the plant defense-signalling hormone SA.
As summarized in Fig. 7, several distinct mechanisms have been adopted by phytopathogens to suppress host SA biosynthesis by cell-entering effectors. The Pseudomonas syringae effector HopI1 remodels host chloroplast thylakoid structure and suppresses SA accumulation via an unknown mechanism 27 . Ustilago maydis secretes a host cytoplasmic effector (Cmu1), which is a chorismate mutase that converts chorismate to prephenate. As chorismate is a key intermediate in the biosynthesis of SA 10 , U. maydis can effectively deplete chorismate and suppress SA accumulation via Cmu1 (ref. 28). Secreted chorismate mutases are found in many plant fungal and nematode pathogens 28,29 . In the current study, we demonstrated that phytopathogens from two kingdoms have evolved a novel strategy to inhibit SA biosynthesis in plants by producing ISC effectors. It is likely to be that neither of the pathogen-generated ISCs are imported into the plastids of host plants, because neither PsISC1 nor VdISC1 were co-localized with the plastid marker pt-rk CD3-999 (ref. 25) when they were transiently co-expressed in the leaves of N. benthamiana (Supplementary Fig. 8). However, when delivered into host plant cells by the pathogens, the ISC effector proteins might be able to affect the cellular homeostasis of isochorismate. Although SA is thought to be synthesized in the plastids, plastid-produced isochorismate may transport to the cytosol where it acts as the substrate of VdIsc1 or PsIsc1 to be catalysed into DDHB. Reduction of cytosolic concentrations of isochorismate by pathogen-derived ISC effectors in turn may enhance its export from the plastid to the cytosol. A similar mechanism was also proposed to explain how U. maydis CMU1 acts on plant chorismate 28 .
Secreted proteins containing ISC family motifs have been found in several fungal and nematode pathogens 16,17 , suggesting a common strategy for host manipulation. Therefore, different phytopathogens have evolved distinct intracellular effectors independently to actively suppress SA accumulation in host cells. The findings suggest a promising strategy for disease control by inhibiting the activities of these enzymatic effectors or supplying the host with plant-defense activators to compensate for the SA shortage caused by pathogen effectors. For example,     ARTICLE several highly active inhibitors targeting the secreted chorismate mutase from Mycobacterium tuberculosis have been designed and evaluated based on structure-guided approaches 30 . Our results suggest that secreted ISCs from phytopathogens could be suitable targets for disease control. We further showed that PsIsc1 and VdIsc1 are secreted proteins, although neither protein contained a predicted signal peptide when we analysed using two widely used algorithms 31,32 . Proteins that lack signal peptides can access the cell exterior via the unconventional protein secretion pathway, which has been documented in animal, plant, bacterial and yeast cells 33,34 . Indeed, both proteins were predicted to be unconventionally secreted proteins using an available algorithm 22 . The secretion mode of ISCs suggests that unconventional secretion occurs in P. sojae and V. dahliae. At least four distinct transport pathways are involved in unconventional secretion 34 , but it remains unknown which (if any) of these mechanisms are used by these two filamentous pathogens. The recruitment of unconventional secretion for effector delivery in both an oomycete and a fungal plant pathogen further broadens the extensive resources pathogens can bring to bear against hosts, and broadens the range of proteins that bioinformaticians must examine as potential effectors.
We used P. sojae to examine the pathway of PsIsc1 secretion and host entry in more detail. We demonstrated that PsIsc1 accumulates predominantly in haustoria and the extrahaustorial matrix during soybean infection, which is a similar distribution to that of well-characterized cell-entering effectors, such as P. sojae Avr1b 19 and P. infestans Avr3a 23 . We did not observe the translocation of fluorescent protein fusions of PsIsc1 inside host cells during infection, which might be due to dilution of the fluorescence signal in plant cell cytoplasm. Nevertheless, we demonstrated that PsIsc1 could functionally replace the N-terminal regions (host target signals, including the signal peptide and RxLR-dEER domain) in Avr1b and mediate the translocation of Avr1b into host cells, consistent with the prediction that PsIsc1 enters into host cells during infection to metabolize the SA precursor. The observation that PsIsc1 was functional when expressed in V. dahliae suggests that either PsIsc1 does not rely on any pathogen-encoded machinery for entry, or that the two pathogens produce functionally analogous translocation machinery.
Assessing how the effectors from filamentous pathogens are translocated into host cells is a major focus of current research. In oomycete pathogen effectors, the RxLR-dEER domain is considered to mediate host cell entry by binding to phosphatidylinositol phosphates on the outer surface of plant plasma membranes 24 . Consistent with this, Phytophthora pathogens could generate external phosphatidylinositol-3-phosphate to promote the entry of RxLR effectors 35 . In contrast, Magnaporthe oryzae intracellular effectors accumulate preferentially in the biotrophic interfacial complex, a novel interfacial structure that is associated with the initial invading hyphae 36,37 , suggesting that the effectors may be translocated through a specialized structure in this fungal pathogen. In the current study, we showed that both PsIsc1 and VdIsc1 probably function inside host cells and PsIsc1 accumulates in haustoria. Nevertheless, the exact mechanism behind the translocation is yet to be investigated.
In summary, these results provide evidence that P. sojae and V. dahliae produce a novel class of effectors: ISC enzymes that destroy the precursor of the plant defense-signalling hormone SA. The discovery of these enzymes in both fungi and oomycetes suggests that this is a central mechanism of virulence in filamentous plant pathogens. We also demonstrated that unconventional secretion has been recruited by plant pathogens in two kingdoms to deliver effectors into host plants; in addition to the biological novelty, this discovery transforms the way that candidate effectors should be identified, which is at present heavily reliant on the identification of signal peptides.

Methods
Bioinformatic analyses. Known ISC protein sequences were obtained from the PDB database (http://www.rcsb.org) 38 and were used as queries to search against the P. sojae 1 and V. dahliae 2 protein sequence databases using reciprocal Blast searches programmes. All the obtained proteins were analysed using HHpred 18 against the PDB and PFAM databases (http://pfam.sanger.ac.uk/) 39 . Proteins were considered to be putative ISCs only if they met the following criteria: in a search of the PFAM database, the proteins were classified as a member of the PF00857 family, which groups together ISCs; and the HHpred searches identified the proteins as belonging to the ISC family (the top ten hits contain over five ISCs). The putative ISCs identified were further confirmed using annotations in the NCBI database.
Microbial strains and plant lines. All of the pathogen strains, the derived transgenic P. sojae and V. dahliae lines, and plant materials used in the current study are summarized in Supplementary Table S3. P. sojae and P. capsici strains were grown and maintained as described previously 35 . V. dahliae strain V991, a highly toxic defoliating WT strain, was grown on PDA medium at 25°C 42 . E. coli and Agrobacterium tumefaciens were grown using appropriate selectable markers. All plants were grown in growth chambers at appropriate temperatures with 16-h photoperiods and 55% humidity.
Generation of transformants and virulence assays. Transformed P. sojae lines were obtained using polyethylene glycol-mediated protoplast transformation procedure 19,35 . Putative transformants were screened by quantitative reverse transcriptase (qRT)-PCR to identify silenced or overexpression transgenic lines. The overexpression lines were confirmed by western blotting using antibodies against the relevant tags. The pathogenicity phenotypes of selected P. sojae transformants were determined by hypocotyl inoculation of etiolated soybean seedlings. Approximately 100 zoospores of each transformant and the WT strain were inoculated onto the hypocotyl of etiolated (susceptible Williams) soybean seedlings. The avirulence phenotypes of selected transformants were evaluated using hypocotyl inoculation 19 with the soybean cultivars Williams (rps) and HARO13 (Harosoy background, Rps1b). Each avirulence or virulence determination was repeated at least three times. Significant differences were identified using Dunnett's test.
V. dahliae transformation was performed using Agrobacterium tumefaciensmediated transformation 43 . Briefly, the conidia (1 Â 10 8 spores per ml) were harvested and then mixed with an equal volume of A. tumefaciens containing the appropriate vector and 200 mM acetosyringone. After incubation, the cultures were transferred to PDA medium containing 50 mg l À 1 hygromycin B (Roche, Switzerland). The putative transformants were screened for gene deletions or the presence of transgenes using genomic PCR and qRT-PCR. For complementation, the mutant strains were further transformed with VdIsc1, PsIsc1 and the different mutants fused with an HA tag. For cotton and Arabidopsis infection assays, plants were inoculated with V. dahliae WT or transformants by conidial root dip 42 . For Arabidopsis inoculations, we modified the dip time to 5 min and the inoculum to 1 Â 10 7 conidia per ml. The symptoms were evaluated and the disease grade was classified as follows: 0 (no symptoms), 1 (40-25% wilted leaves), 2 (25-50%), 3 (50-75%) and 4 (75-100%). The data were analysed using Student's t-test.
Molecular experiments. Promoters, flanking sequences, complete open reading frames and/or fragments of PsIsc1, VdIsc1, Avr1b and EntC were amplified by PCR using specific primers (Supplementary Table 4) and cloned via standard restriction digest and ligation methods to produce expression constructs based on the vectors pC1300, pTOR, pET21b, pBinGFP2 or pBinGFP4 (Supplementary Table 5).
Total RNA was extracted from cotton roots using the CTAB method 42 . Total RNA was extracted from N. benthamiana leaves, etiolated soybean seedlings, and P. sojae and V. dahliae mycelium using RNA simple Total RNA Kit (Tiangen, China) following the manufacturer's instructions. qRT-PCR was performed using SYBR green real-time RT-PCR assays (TaKaRa) as described previously 35 . To assess the transcriptional profile of genes from P. sojae, total RNA was collected from P. sojae P6497 mycelium sandwiched between pairs of soybean leaves. The mycelium was harvested at 2, 4 and 10 h after inoculation (hpi) and the infected soybean leaves, including mycelia, were collected at 12, 24, 36 and 48 hpi. To assess the transcriptional profile of genes from V. dahliae inoculum, mycelia were inserted into wounded N. benthamiana petioles and the mycelia were recovered at 6, 12, 24 hpi, and the infected petioles, including mycelia, were also collected at 3 and 5 dpi.
Western blotting. Protein extracts were prepared by grinding B100 mg leaf or slightly dried mycelia samples in liquid nitrogen. The resulting powder was then suspended in protein extraction buffer (50 mM HEPES, 150 mM KCL, 1 mM EDTA, 0.1% Triton X-100, pH 7.5) supplemented with 1 mM dithiothreitol and 1 Â protease inhibitor cocktail (Roche). The resultant suspensions were mixed and centrifuged at 12,000g at 4°C for 15 min and the supernatants were recovered. Samples were fractionated using standard SDS-PAGE gels (12% acrylamide) and blotted using standard procedures 44 . Anti-His-tag primary monoclonal antibody (Sigma-Aldrich) and IRDye 800CW-conjugated goat (polyclonal) anti-mouse IgG (H þ L; LI-COR Biosciences) secondary antibodies were used to assess P. sojae transformants. Anti-GFP monoclonal antibody (Sigma-Aldrich) and IRDye 800CW-conjugated goat anti-mouse IgG secondary antibodies were used to assess transient gene expression in plants. All the primary antibodies were diluted with 5% (m/V) non-fat dried milk in PBS (1:3,000) and the secondary antibodies were diluted with 5% (m/V) non-fat dried milk in PBS (1:10,000). The membranes were then visualized using a LI-COR Odyssey scanner with excitation at 700 and 800 nm. The molecular marker is shown in Supplementary Fig. 6 and blots were shown with at least one size-labelled marker.
V. dahliae strain D1-12VdIsc1, P. sojae strains OT20, PsIsc1-mRFP, and Avr1b-mRFP and GFP were used to analyse the secretion of VdIsc1 and PsIsc1. The proteins were purified from the same amounts of culture (including the mycelia and supernatants) for each strain. For P. sojae, mycelia were cultivated in flasks at 25°C with shaking at 100 r.p.m. for 2 weeks in liquid Czapek culture medium. Culture supernatants were precipitated using ammonium sulphate and then analysed using western blotting with mouse anti-His (Sigma-Aldrich) and anti-aactin antibodies (Sigma, A3853). For V. dahliae, mycelia were cultivated in a flask for 2 weeks at 25°C in liquid Czapek-Dox medium and the secreted proteins were collected following the reported procedures 45 . Culture supernatants were precipitated using trichloroacetic acid and then analysed using western blotting with mouse-anti HA (Sigma, H3663) and anti-a-actin antibodies.
N. benthamiana infiltration and virulence assays. A. tumefaciens strain GV3101 containing the appropriate vector was infiltrated into N. benthamiana leaves 35 . P. capsici infection assays were performed using droplet inoculations of zoospore solutions of P. capsici isolate 35 (10 ml, B500 zoospores) onto detached N. benthamiana leaves. At least ten independent (4-week-old) N. benthamiana leaves were tested per construct combination. To explore the effect of transient expression on infection, proteins were first expressed by infiltrating cells of the relevant A. tumefaciens strain into N. benthamiana leaves 24 h before P. capsici inoculation. The growth efficiency of P. capsici was quantified by measuring the lesion diameter (mm) at 48 hpi.
Measurement of SA/SAG concentrations and application. SA was extracted from leaves excised from soybean, cotton and N. benthamiana plants, and quantified using HPLC as described 46 with minor modifications. Briefly, leaf powder (B1.0 g) was ground in liquid nitrogen and then suspended in 90% (v/v) methanol. As an internal standard for SA, 100 mg 3-hydroxy benzoic acid in 100% methanol were added to each sample. The SA solution was filtered and separated on a C18 analytical column using HPLC and detected using fluorescence (excitation at 305 nm, emission at 405 nm; Waters). The HPLC was programmed for isocratic conditions with a flow rate of 0.5 ml min À 1 . To assay SAG, the aqueous phase containing SAG was acidified to pH 1.0 using HCl and was then boiled for 30 min to release SA from any acid-labile conjugated forms. The released SA was extracted using an organic mixture and treated as described above. SA and SAG were quantified by area integration of the HPLC peaks.
Cotton plants were sprayed with 2.0 mM SA 1 day before inoculation or were mock treated with water before infection. Following inoculation, the plants were sprayed with SA every 7 dpi.
Enzyme activity assays and measurement of DDHB content. Isochorismatase activity was measured as described previously 13 . Briefly, the enzyme EntC was used to convert chorismate to isochorismate and enzyme activity was then determined by measuring the increase in absorbance at 340 nm caused by the production of NADH from NAD þ when V. cholerae VibA was present in the reaction solution (converting DDHB þ NAD þ to 2,3-DHB þ NADH) using an Uvikon spectrophotometer. Purified E. coli EntB and V. cholerae VibB produced from E. coli were used as positive controls 13 . The tested proteins were purified from N. benthamiana transiently expressing VdIsc1 or PsIsc1 genes using HisTrap HP (GE Healthcare) following the manufacturer's instructions. The assay was performed in a total volume of 100 ml containing 10-ml reaction solutions from the first step, 100 mM PBS buffer pH 7.0, 0.8 mM NAD þ , 4 mg of each purified protein and 3.8 mg of VibA at room temperature.
To measure the relative DDHB concentration in plant tissues, the crude extracts were collected from B1 g N. benthamiana leaves. The assay was performed in a total volume of 100 ml containing 100 mM PBS, pH 7.0, 10-ml crude extracts, 3.8 mg purified VibA, 0.8 mM NAD þ , 100 mM PBS buffer pH 7.0 at room temperature for 10 min. Next, the absorbance at 340 nm owing to NAD þ reduction was used to calculate the relative DDHB contents in the tested samples.
Microscopy. One hundred zoospores from each P. sojae transformant were inoculated onto the hypocotyl of the susceptible soybean Williams etiolated seedlings. The surface layer of the inoculated tissue was then examined B10 hpi. GFP-VdIsc1 and GFP-PsIsc1 fusions were expressed together with the mCherrylabelled plastid localization marker (pt-rk CD3-999) 25 in N. benthamiana leaves. Protein fluorescence in the N. benthamiana leaves was observed using a ZEISS LSM 710 confocal microscope (Zeiss Microsystems) with a Â 40 objective lens (Zeiss) at the specific excitation and emission wavelengths (RFP, 561 and 600-650 nm; GFP, 488 and 495-530 nm; mCherry, 587 and 600-650 nm).