• The avirulence gene ACE1 from the rice blast fungus Magnaporthe grisea encodes a polyketide synthase (PKS) fused to a nonribosomal peptide synthetase (NRPS) probably involved in the biosynthesis of a secondary metabolite recognized by Pi33 resistant rice (Oryza sativa) cultivars.
• Analysis of the M. grisea genome revealed that ACE1 is located in a cluster of 15 genes, of which 14 are potentially involved in secondary metabolism as they encode enzymes such as a second PKS-NRPS (SYN2), two enoyl reductases (RAP1 and RAP2) and a putative Zn(II)2Cys6 transcription factor (BC2).
• These 15 genes are specifically expressed during penetration into the host plant, defining an infection-specific gene cluster. A pORF3-GFP transcriptional fusion showed that the highly expressed ORF3 gene from the ACE1 cluster is only expressed in appressoria, as is ACE1. Phenotypic analysis of deletion or disruption mutants of SYN2 and RAP2 showed that they are not required for avirulence in Pi33 rice cultivars, unlike ACE1. Inactivation of other genes was unsuccessful because targeted gene replacement and disruption were inefficient at this locus.
• Overall, the ACE1 gene cluster displays an infection-specific expression pattern restricted to the penetration stage which is probably controlled at the transcriptional level and reflects regulatory networks specific to early stages of infection.
Polyketides are major fungal secondary metabolites with diverse biological activities (Manzoni & Rollini, 2002) that have important roles in pathogenicity on plants, mainly as phytotoxins (Yang et al., 1996; Choquer et al., 2005), and impacts on livestock and human health (Bennett & Klich, 2003). These metabolites are synthesized by polyketide synthases (PKSs), which are large enzymes related to fatty acid synthases (FASs; Cox, 2007). Characterization of these biosynthetic pathways showed that additional enzymes are required for the modification of the precursor synthesized by the PKS (Hoffmeister & Keller, 2007). The genes that encode PKS and modifying enzymes from the same pathway are typically located at the same locus and co-expressed, defining a gene cluster (Keller & Hohn, 1997). Secondary metabolism gene clusters involved in polyketide biosynthesis were identified in a large number of fungi (Hoffmeister & Keller, 2007), including plant pathogens such as Gibberella spp. (fumonisin/zearalenone; Proctor et al., 2003; Kim et al., 2005) and Cercospora spp. (cercosporine; Chen et al., 2007). Genes encoding modifying enzymes such as oxidoreductases, P450 monooxygenases, O-methyl transferases, ketoreductases and esterases are located in these clusters (Brown et al., 1996; Kennedy et al., 1999; Proctor et al., 2003; Yu et al., 2004; Kim et al., 2005). These clusters also include genes encoding ATP-binding cassette (ABC) or major facilitator superfamily (MFS) transporters whose role remains elusive or limited to self protection as their inactivation does not always affect the efflux of the polyketide (Proctor et al., 2003; Chang et al., 2004), or affects it only quantitatively (Choquer et al., 2007). Expression of genes in clusters is generally controlled by a specific transcriptional activator encoded by a gene located in the cluster. For example, LovE and LovH in the lovastatin cluster are specifically involved in the regulation of the production of lovastatin (Hutchinson et al., 2000), while aflR, ZEB2 and CTB8 control the expression of genes in the sterigmatocystin/aflatoxin (Woloshuk et al., 1994), zearalenone (Kim et al., 2005) and cercosporin (Chen et al., 2007) clusters, respectively. Transcriptional regulation of the fumonisin gene cluster is more complex as it requires two transcription factors from the Zn(II)2Cys6 binuclear family, one located in the cluster (FUM21) and another (ZFR1) located elsewhere in the genome (Flaherty & Woloshuk, 2004; Brown et al., 2007).
The avirulence (AVR) gene ACE1 from the rice blast fungus Magnaporthe grisea differs from other fungal AVR genes because it encodes a hybrid polyketide synthase/nonribosomal peptide synthetase (PKS-NRPS) that is not secreted (Böhnert et al., 2004). Thus, Ace1 is an enzyme that is probably involved in the biosynthesis of a polyketide. Magnaporthe grisea isolates or transformants carrying the functional AVR gene ACE1 are unable to infect rice (Oryza sativa) cultivars carrying the corresponding R gene Pi33, while isolates or mutants defective for ACE1 bypass the rice Pi33-mediated resistance and infect such resistant cultivars (Böhnert et al., 2004). Modification of an essential amino acid of the catalytic site of the β-ketoacyl synthase domain of Ace1 abolishes avirulence, although the mutant protein is still produced in appressoria. As Ace1 biosynthetic activity is required for avirulence, the avirulence signal recognized by Pi33 rice cultivars is not the Ace1 protein, but the secondary metabolite synthesized by Ace1 (Böhnert et al., 2004). ACE1 is only expressed in mature appressoria differentiated on plant leaves or artificial cellophane membranes during the penetration process (maximum of expression 17–24 h after inoculation; Fudal et al., 2007). ACE1 is not expressed in appressoria formed on artificial membranes that cannot be pierced (Teflon) or in penetration-defective appressoria of the melanin-deficient mutant buf1 (Fudal et al., 2007). Similarly, inhibition of the penetration process by treatment of mature appressoria with actin or tubulin inhibitors abolishes ACE1 expression (Fudal et al., 2007). These experiments demonstrate that ACE1 appressorium-specific expression does not depend on host plant signals, but is connected to the onset of appressorium-mediated penetration. Here, we report that ACE1 belongs to a cluster of secondary metabolism genes specifically expressed at an early stage of the infection. This cluster contains 15 genes encoding two PKS-NRPSs (ACE1 and SYN2), 10 enzymes involved in polyketide modification such as two enoyl reductases (RAP1 and RAP2), a protein with unknown function (ORF3), a MFS transporter (MFS1) and a putative transcription factor (BC2). Gene inactivation of SYN2 and RAP2 showed that they are not required for avirulence in Pi33 resistant rice cultivars.
Materials and Methods
Fungal strains, growth conditions and transformation
The avirulent fertile isolate Guy11 which is pathogenic on rice (Nottéghem et al., 1994), the Guy11-ACE1(p)-GFP transformant (Fudal et al., 2007) and the Guy11-Δku80 strain (Villalba et al., 2008) were used in this study. Fungal strains were grown and stored as described by Dioh et al. (2000). Protoplasts of Guy11 and Guy11-Δku80 were prepared as described previously (Böhnert et al., 2004) and transformed following the procedure of Villalba et al. (2008). Agrobacterium tumefaciens-mediated transformation was performed as described by Rho et al. (2001). For glufosinate selection, transformants were selected on the complex medium defined by Sweigard et al. (1997) (1.6 g l−1 Difco yeast nitrogen base (YNB) without amino acid, 2 g l−1 asparagine, 1 g l−1 NH4NO3 and 10 g l−1 glucose adjusted to pH 6) containing 35 mg l−1 glufosinate (Cluzeau Info Labo, Ste Foy la Grande, France). For hygromycin selection, transformants were selected on Tanaka complete medium (10 g l−1 glucose, 2 g l−1 yeast extract, 2 g l−1 NaNO3, 2 g l−1 KH2PO4, 0.5 g l−1 MgSO4·7H2O, 0.1 g l−1 CaCl2·2H2O, 0.004 g l−1 FeSO4·7H2O and microelements, adjusted to pH 5.5–5.8) containing 120 mg l−1 hygromycin (Sigma-Aldrich, St Louis, MO, USA). Transformants were purified by isolation of a single spore. Strains were grown on liquid Tanaka complete medium for genomic DNA extraction.
Nucleic acid extractions
Genomic DNA was isolated from Magnaporthe grisea Hebert (Barr) mycelium following the phenol/chloroform protocol described by Sweigard et al. (1995) with lyophilized ground mycelium serving as starting material instead of protoplasts. Total RNA was extracted from detached barley leaves infected by the Guy11 wild-type strain at 0, 8, 17, 24, 30, 48, 52 and 72 h after inoculation (see the Pathogenicity assays section for infection procedure). Infected plant tissue was cut and frozen in liquid nitrogen, followed by lyophilization overnight. Cut lesions were ground in a mortar with liquid nitrogen and total RNA was extracted using a hot acid phenol protocol. A fine powder (approx. 2 ml) was incubated at 65°C for 1 h in 3 ml of TES (10 mm Tris-HCl, pH 7.5, 10 mm ethylenediaminetetraacetic acid (EDTA) and 0.5% sodium dodecyl sulphate (SDS)) and 3 ml of acid phenol (pH 4.8; Sigma-Aldrich). Samples were incubated for 5 min on ice before a 15-min centrifugation (6000 g, 4°C). The aqueous phase was washed twice with an equivalent volume of acid phenol following the same procedure (vortexing, incubation on ice and centrifugation), followed by one wash with chloroform. Total RNA was then precipitated with 3 M sodium acetate, pH 5.2 (1 : 10 volume), and cold ethanol (2.5 volume). Samples were subjected to centrifugation for 10 min at 4°C (13 000 g) and the pellet was washed with cold ethanol (70%). After centrifugation, the pellet was dried and made soluble in 100 µl of nuclease-free water.
Genomic DNA libraries of 96/0/76 and parental Guy11 strains were constructed by Dioh et al. (1997). These libraries were hybridized with single radiolabeled probes (50 ng of PCR products, plasmids or digested cosmids corresponding to sequences from ACE1, CYP2 and SYN2, labeled using the Amersham MegaPrime kit following the manufacturer's protocol; Amersham Biosciences, Little Chalfont, UK). Hybridizing clones were grown in Luria-Bertoni (LB) liquid medium and cosmid DNA isolated using the DNA Maxi preps kit (Qiagen, Germantown, MD, USA). Cosmids D31C12, D22F11 and D16C4 were retrieved from the 96/0/76 library, and cosmid G22D7 from the Guy11 library.
All primers used in this study are available as Supplementary Material. Reverse transcriptase–polymerase chain reaction (RT-PCR) and 5′ and 3′ rapid amplification of cDNA ends (RACE) experiments were carried out on total RNA extracted from 17-h infected detached barley (Hordeum vulgare) leaves. The transcription initiation and polyadenylation sites of BC2 were determined by 5′ and 3′ RACE using the GeneRacerTM kit (Invitrogen, Carlsbad, CA, USA) and sequence analysis of amplification products. 5′ and 3′ untranslated regions (UTRs) were amplified with the gene-specific primers 5UTRBC2 and 3UTRBC2, respectively. For intron characterization, RT-PCR were carried out with 5 µg of total RNA as starting material and suitable primers using ReadyToGo™ You-Prime First-Strand Beads (Amersham Biosciences) according to the manufacturer's protocol. Amplification products were sequenced.
PCR analyses of transformants were carried out using 50 ng of DNA template in 50-µl amplification reactions that contained sets of two primers targeting the genes of interest, their flanking regions or deletion cassettes. Thirty cycles of PCR were performed and DNA products were analyzed on 0.8% agarose gels. Southern blot analyses were carried out as described by Villalba et al. (2008).
The reverse transcriptase reaction was carried out using the ThermoscriptTM RT-PCR system (Invitrogen) with 5 µg of total RNA extracted from infected barley leaves at several time-points and from mycelium grown in complete liquid agitated medium for 24 h (three biological replicates for each time-point). Primers were designed using Primer Express® software (Applied Biosystems, Foster City, CA, USA). Real-time PCR was carried out using a LightCycler 1.0 (Roche Diagnostics, Indianapolis, IN, USA) using the Fast Start DNA Master SYBR® Green I kit (Roche Diagnostics). Expression data were normalized for a given time-point according to the formula 2−ΔCt = 2−(Ctgene X – Ctgene ref), where Ct is the cycle threshold, with ILV5 as the reference gene, and expression relative to ACE1 was calculated according to the formula 2−ΔΔCt = 2−[(Ctgene X – Ctgene ref) – (CtACE1 – Ctgene ref)], with ILV5 as the reference gene (Livak & Schmittgen, 2001). StatBox version 6.5 (Grimmersoft, Paris, France) was used for statistical analyses (critical threshold α = 0.05).
Cosmids and plasmids were sequenced by Genome Express (Meylan, France). Vector NTI v9 (Invitrogen) was used to define open reading frames. Protein alignment was performed with ClustalW (Thompson et al., 1994) and edited in GeneDoc version 2.6.002 (K.B. Nicholas & H.B. Nicholas, unpublished). FgenesH (http://www.softberry.com) was used to identify introns using Schizosaccharomyces pombe and Neurospora crassa matrices. Search for homologies and protein conserved domains was carried out using the BLASTP algorithm against the fungal database (http://www.ncbi.nlm.nih.gov) and using the InterProScan tool (http://www.ebi.ac.uk).
For cloning purposes, the Escherichia coli DH5α strain was used. Molecular methods followed standard protocols (Sambrook et al., 1989). The ORF3 (MGG_08381.5) terminator (1589 bp) was amplified with primers TERMORF3-NotI+ and TERMORF3-EcoRI– using Pfu Turbo polymerase (Invitrogen). The PCR product, digested with NotI and EcoRI, was introduced into plasmid pCB1635, which was digested with the same restriction enzymes using the Rapid DNA ligase kit (Roche Diagnostics), resulting in plasmid pCB1635-TERMORF3, which carries a glufosinate resistance marker. The ORF3 promoter, corresponding to 378 bp upstream of ATG, was synthesized by GenScript Corp. (Piscataway, NJ, USA), with a NotI restriction site being inserted on the 5′ side and a NcoI site at the ATG (pUC57-ORF3(p)). The reporter gene GFP was obtained by NcoI and SnaBI digestion of plasmid promACE1:eGFP (Fudal et al., 2007), and the 1484-bp fragment obtained was inserted into pUC57-ORF3(p), which was digested by NcoI and SmaI. The ORF3 promoter linked to GFP was retrieved by NotI restriction and introduced into pCB16335-TERMORF3, which was linearized by NotI, resulting in plasmid pORF3(p)-reporter.
The sequences flanking SYN2, SYNL (998 bp) and SYNR (1236 bp), were amplified using Taq DNA polymerase (Q-BIOgene, Irvine, CA, USA), with cosmid D16C4 as template. The SYNR PCR product and plasmid pBSK were digested with HindIII and EcoRI, followed by ligation with the Rapid DNA ligase kit (Roche Diagnostics). SYNL was cloned into the pGEM®-T Easy Vector (Promega, Madison, WI, USA), released by EcoRI and introduced into the plasmid pBSK-HygR, which contains a hygromycin resistance cassette. A 2.4-kb fragment corresponding to SYNL and linked to the hygromycin resistance cassette was released by PstI and HindIII digestion and ligated with the SYNR HindIII-EcoRI fragment and plasmid pBSK, which was digested by PstI and EcoRI using T4 DNA Ligase (Roche Diagnostics). Finally, this intermediate plasmid was digested with PstI and EcoRI and the 3.4-kb deletion cassette obtained was introduced into the pDHt vector (Mullins et al., 2001). The resulting plasmid was transformed into the A. tumefaciens AGL-1 strain (Mullins et al., 2001) by electroporation as described by Rho et al. (2001).
An internal fragment of RAP2 (971 bp) with KpnI and XbaI restriction sites at its 5′ and 3′ ends, respectively, was amplified using Taq DNA polymerase (Q-BIOgene), with cosmid D16C4 as template. The agarose gel purified fragment was cloned using the TOPO-TA cloning kit (Invitrogen). The resulting vector and plasmid pCB1003 (Sweigard et al., 1997), which contains a hygromycin resistance cassette, were digested with KpnI and XbaI, and ligated overnight using T4 DNA Ligase (Roche Diagnostics).
Pathogenicity assays on barley were carried out with 10-d-old seedlings of cultivar Plaisant cultivated at 20°C during the day, and 15°C during the night. Fungal spores were harvested from 11- to 14-d-old M. grisea rice agar culture (20 g l−1 rice powder, 10 g l−1 glucose, 2 g l−1 KH2PO4, 3 g l−1 KNO3 and 15 g l−1 agar, pH 6). Detached leaf assays were carried out as follows: 35 µl of calibrated spore suspensions (3 × 105 spores ml−1) were deposited onto leaves placed on water agar medium (3%) containing kinetin (2 µg ml−1, to keep the detached leaves alive longer). Plates were incubated at 26°C under a photoperiod of 12 : 12 h light:dark. Lesions were collected at 0, 8, 17, 24, 30, 48, 52 and 72 h after inoculation.
Pathogenicity assays on rice were carried out with seedlings from susceptible cultivars Sariceltik and CO39, and Pi33 resistant rice cultivars C101LAC and Bala (four-leaf growth stage). One day after, 10 ml of calibrated spore suspensions (3 × 104 spores ml−1 containing 0.3% gelatin) was sprayed onto each pot, containing approx. 10 plants, with the addition of 15 ml ammonium sulfate (12 g l−1, to increase the susceptibility of rice to M. grisea). Inoculated rice plants were incubated in a humid chamber at 100% humidity for 24 h at 25°C in darkness, and subsequently transferred to a glasshouse at 25°C and 50% humidity under a photoperiod of 12 : 12 h light:dark. Symptoms were observed 6 d after infection.
Characterization of the genes from the ACE1 locus
The avirulence gene ACE1 was identified by map-based cloning at one end of M. grisea cosmid clone D31C12 (Böhnert et al., 2004). This cosmid contains seven other open reading frames (ORFs) without homology to known proteins, two transposable elements and a gene (MGG_08401.5) encoding a putative xylanase. Additional overlapping cosmids (D22F11, D16C4 and G22D7; Fig. 1) obtained from a chromosome walk span a genomic region of approx. 180 kb including the D31C12 end containing ACE1. We have determined the nucleotide sequence of the 75 kb upstream of ACE1. This region contains 20 putative genes (MGG_01905.5 to MGG_08391.5; Fig. 1, Table 1) and four copies of transposable elements (Pot2 and Pot3-Mg-SINE). Introns were predicted using ab initio prediction software FgenesH (http://www.softberry.com), by searching for intron splicing site consensus sequences and by alignments with the most similar fungal proteins. Removal of putative introns corrected reading frame shifts in the ORFs and defined coding sequences (CDSs). This annotation was confirmed for some genes by sequencing partial cDNAs obtained either by RT-PCR of intron–exon junctions (MGG_08380.5, MGG_08386.5 and MGG_08391.5) or by 5′/3′ RACE (MGG_08386.5).
Table 1. Predicted function of genes in the ACE1 cluster
The first BLASTP match is the closest homolog, consistent with the phylogenetic analysis of Khaldi et al. (2008). Only one homolog is indicated when the first BLASTP match is also the match with known function.
Conserved domains were identified using InterProScan. KS, β-ketoacyl synthase (IPR000794); AT, acyl transferase (IPR001227); MT, methyltransferase (IPR013217); KR, ketoreductase (IPR013968); ACP, acyl carrier protein (IPR006163); C, condensation domain (IPR001242); A, AMP-dependent synthetase and ligase (IPR000873) corresponding to the adenylation domain; PP, phosphopantetheine binding (IPR006163).
Transposons were identified using the BLASTN algorithm.
The genomic sequence of the ACE1 locus was obtained from M. grisea isolate Guy11. However, M. grisea publicly available sequence (http://www.broad.mit.edu) was obtained from strain 70-15, which is derived from Guy11 through numerous outcrosses and backcrosses (Leung et al., 1988; Chao & Ellingboe, 1991) and has a functional ACE1 gene as this strain is also avirulent in Pi33 rice resistant cultivars (D. Tharreau, pers. comm.). Several modifications of available predictions deduced from the genome sequence of the 70-15 strain were performed. The main modifications involved the division of MGG_08385.5 into two genes (CYP2 and OXR2), and the fusion of MGG_12451.5 and MGG_12452.5 into a single gene, SYN2 (Table 1, Supplementary Material Fig. S1). Additionally, ORF and/or intron predictions from the 70-15 genome sequence were slightly modified for seven predicted genes (MGG_08379.5, MGG_08384.5, MGG_08385.5, MGG_08387.5, MGG_08389.5, MGG_12447.5 and MGG_12451.5/MGG_12452.5; Table 1, Table S1) according to the Guy11 sequence. Overall, the comparison of sequences from the ACE1 locus in isolates Guy11 and 70-15 revealed only a few differences (94.8% identity) although some had a dramatic effect (Supplementary Material Fig. S1). The MGG_08386.5 (BC2) coding sequence from 70-15 has a single nucleotide insertion compared with Guy11, leading to a frame shift (947 bp from the start codon, which corresponds to the end of the ORF). However, this mutation only shortens the Bc2 protein by 33 amino acids, suggesting that it is still functional because it does not affect its predicted functional domain. MGG_08385.5 (CYP2), MGG_12451-2.5 (SYN2) and MGG_08384.5 (MFS1) have, respectively, a 636-bp deletion in the coding sequence (594 bp from the start codon), an insertion of a single base pair leading to an early stop codon (4436 bp from the start codon), and a deletion of a single base pair leading to an early stop codon (687 bp from the start codon). These mutations probably inactivate the CYP2, MFS1 and SYN2 genes. The MGG_08387.5 sequence from 70-15 has a 48-bp insertion compared with Guy11 which does not change the protein as it is located in the fourth predicted intron. Sequences from other genes are identical between the two strains. Most of the genes from the Guy11 ACE1 locus correspond to genes in 70-15 that are located on chromosome I as predicted from the genetic mapping of ACE1 on chromosome I (Dioh et al., 1997). However, three genes (MGG_01903.5 to MGG_01905.5) at the upstream end of the sequenced region of the Guy11 ACE1 locus correspond to genes in 70-15 that are located on chromosome II. As the G22D7 cosmid and Guy11 genomic sequences have the same restriction pattern, this linkage suggests that Guy11 and 70-15 differ by a recombination event between chromosomes I and II near the ACE1 locus (Fig. 1).
BLASTP analysis and a search for known or conserved domains revealed that 14 genes from the ACE1 locus encode proteins with significant similarities to enzymes or proteins involved in the biosynthesis of fungal secondary metabolites and its control (Table 1). The remaining genes encode predicted or hypothetical proteins without known functions. These 14 genes potentially involved in secondary metabolism are located in a 71-kb region spanning ACE1 to MGG_08377.5 (OME1). OME1 encodes a putative SAM-dependent O-methyl transferase highly similar to omtB from Aspergillus flavus, which is involved in aflatoxin biosynthesis (Yu et al., 2000). In this region, the intergenic distances vary from 435 to 3136 bp (average 1.4 kb). Located 42 kb upstream of ACE1 and 1.8 kb downstream of OME1, SYN2 encodes a hybrid PKS-NRPS highly similar to Ace1 and the lovastatin nonaketide synthase (LNKS) from Aspergillus terreus. Enzymatic domains corresponding to β-ketoacyl synthase (KS), acyl transferase (AT), keto-reductase (KR), dehydratase (DH), methyl transferase (MT) and acyl carrier protein (ACP) were identified in the Syn2 protein sequence, as well as condensation (C), adenylation (A) and thiolation (peptidyl carrier protein PCP) domains corresponding to a single NRPS module. We also identified in this region RAP1 and RAP2, which encode proteins highly similar to the A. terreus enoyl reductase LovC which is involved in the biosynthesis of the polyketide lovastatin (Kennedy et al., 1999). Four cytochrome P450 monooxygenase encoding genes, CYP1, CYP2, CYP3 and CYP4, and two putative oxidoreductase genes, OXR1 (short-chain alcohol dehydrogenase) and OXR2 (FAD-dependent oxidase), were also identified between ACE1 and SYN2. Oxr2 is related to Zeb1 from Gibberella zeae, which oxidizes β-zearalenonol to zearalenone (Kim et al., 2005). The gene MFS1 identified between CYP2 and RAP2 encodes a putative transporter from the major facilitator superfamily that could play a role in polyketide efflux and self-protection. The gene BC2 located between CYP1 and OXR2 encodes a putative transcriptional regulator with a Zn(II)2Cys6 binuclear cluster DNA-binding domain. Finally, two other genes, ORF3 and ORFZ, were identified between RAP1 and RAP2. ORF3 has no homology to known proteins while ORFZ shares homologies with alpha/beta-hydrolases. Overall, the characterization of the ACE1 locus led to the identification of 15 co-linear genes, with 14 of them showing significant similarities to known genes involved in secondary metabolism. This putative cluster is flanked by genes encoding predicted or hypothetical proteins.
Expression of genes from the ACE1 locus
ACE1 is specifically expressed in appressoria during the penetration of the fungus into host leaves (Böhnert et al., 2004; Fudal et al., 2007). The expression of the 14 genes located upstream of ACE1 and ACT1 was monitored using real-time RT-PCR and normalized to ILV5 expression level. RNA was isolated from fungal mycelium grown in culture (complete medium), from infected barley leaves at different time-points of infection, and from healthy barley leaves. Genes in the putative ACE1 cluster were not expressed in mycelium as observed for ACE1, with the exception of OXR1, which displayed significant, although very low, expression in mycelium (2−ΔCt = 4.7 × 10−3; 600-fold below that of ACT1 encoding actin: 2−ΔCt = 2.8). These 14 genes showed the highest expression 17 h post inoculation (hpi) as observed for ACE1 (Fig. 2). Their expression levels slightly decreased at 24 and 30 hpi and fell dramatically at 48 h. The level of expression at 17 hpi (maximum) varied considerably among genes. The ACE1 expression level (2−ΔCt = 1.0) was very similar to that of the actin-encoding gene ACT1 (2−ΔCt = 1.6). ORF3 and ORFZ displayed expression levels 7- and 2-fold higher than that of ACE1, respectively (Fig. 2a). The expression levels of RAP1, RAP2, CYP1 and CYP4 were similar to that of ACE1. The OXR2 expression level was half that of ACE1, while those of SYN2, OXR1, CYP2 and CYP3 were 3- to 7-fold lower than that of ACE1 (Fig. 2b,c). The lowest expression levels (15- to 30-fold lower than that of ACE1) were observed for BC2, MFS1 and OME1 (Fig. 2c). Monitoring of the expression of the genes located downstream of ACE1 (MGG_08375.5 and MGG_08376.5) and upstream of OME1 (MGG_08394.5 and MGG_08395.5) revealed that they were not expressed in mycelium or in infected leaves. Therefore, these genes that encode proteins with unknown function do not share the infection-specific expression pattern of genes from the ACE1 cluster.
The expression pattern of ACE1 determined by quantitative RT-PCR correlates well with its appressorium-specific expression restricted to the penetration stage observed using pACE1-eGFP transcriptional fusion (Fudal et al., 2007). To confirm that genes in this cluster are specifically expressed in the appressorium during infection, we constructed an expression vector with eGFP under control of the ORF3 promoter and terminator sequences. This ORF3 expression vector was introduced by transformation into the Guy11 wild-type isolate. The mycelium and spores of the eight transformants analyzed did not display any eGFP fluorescence. These transformants displayed a medium to strong appressorium-specific eGFP fluorescence during the infection of barley leaves (Fig. 3), as observed for ACE1 (Fudal et al., 2007). These experiments strongly suggest that ACE1 belongs to a cluster of 15 secondary metabolism genes with the same infection-specific expression pattern restricted to appressoria during the penetration of the fungus into host plant tissues.
Functional analysis of genes from the ACE1 cluster
ACE1 encodes a hybrid PKS-NRPS that is involved in the biosynthesis of a secondary metabolite that is probably recognized by rice cultivars carrying the resistance gene Pi33 (Böhnert et al., 2004). As we have shown that ACE1 belongs to a cluster of secondary metabolism genes specifically co-expressed at an early stage of infection, we have initiated a functional analysis of the 14 genes in this cluster to evaluate their role in avirulence in Pi33 resistant rice cultivars. To address this question, we first constructed SYN2 null mutants as this gene encodes the second PKS-NRPS of the cluster. A targeted gene replacement vector pBSK::SYN was constructed to replace a 1.2-kb internal fragment of SYN2, which contains part of its promoter and first exon corresponding to its KS domain, with a hygromycin B resistance cassette (Supplementary Material Fig. S2). This vector was introduced into the Guy11 wild-type strain using A. tumefaciens-mediated transformation. Two transformants (numbers 134 and 140) among 46 tested had a PCR pattern corresponding to the replacement of SYN2 by the Δsyn2::hph allele and were confirmed by Southern analysis (Supplementary Material Fig. S2). These two Guy11-Δsyn2 deletion mutants were inoculated onto Pi33 resistant rice cultivar C101LAC and susceptible rice cultivar CO39 (Fig. 4). The avirulent parental strain Guy11, the virulent strain corresponding to the deletion mutant of ACE1 (Guy11-Δace1) and transformant 136 resulting from the ectopic integration of pBSK::SYN were used as controls. The virulent strain Guy11-Δace1 was able to infect both cultivars while the avirulent strain Guy11, the ectopic transformant 136 and the two Guy11-Δsyn2 mutants were fully pathogenic on CO39 but unable to infect resistant cultivar C101LAC. These results demonstrate that SYN2 is not required for the biosynthesis of the avirulence signal recognized by resistant rice cultivars carrying Pi33.
Attempted deletion of other genes from the ACE1 cluster using targeted gene replacement failed, even in the Guy11-Δku80 background, which increases the gene targeting frequency to 98% (Villalba et al., 2008). To circumvent this problem, we performed gene disruption experiments to insert a resistance cassette into genes from the ACE1 cluster in both Guy11 wild-type and Guy11-Δku80 backgrounds. To this end, an internal sequence of RAP2 was cloned into the vector pCB1003 which carries a hygromycin B resistance cassette (Supplementary Material Fig. S3). This vector was introduced into M. grisea protoplasts using a polyethylene glycol (PEG)-mediated transformation protocol. Disruption mutants were not obtained from Guy11 transformation for 43 transformants tested. Two transformants (numbers 18 and 28) of 40 obtained in a Guy11-Δku80 background displayed a pattern characteristic of the integration of the vector into RAP2. Southern blot analysis revealed that these integration events were complex for both mutants, resulting in tandem integration of the plasmid into RAP2 (Supplementary Material Fig. S3). Pathogenicity and avirulence assays were carried out on Pi33 resistant rice cultivar C101LAC and susceptible rice cultivar CO39 (Fig. 4). Avirulent parental strain Guy11, deletion mutant Guy11-Δace1 and transformant 2, which carries an ectopic integration of pCB1003::RAP2, were used as controls. Guy11, ectopic transformant 2 and putative rap2::hph mutants were unable to infect resistant cultivar C101LAC, while they were fully pathogenic on susceptible cultivar CO39. These results show that RAP2 is not required for the biosynthesis of the avirulence signal recognized by resistant rice cultivars carrying Pi33. However, the possibility cannot be excluded that the closely related enoyl reductase Rap1 could compensate for the loss of RAP2.
Characterization of the ACE1 secondary metabolism gene cluster
Polyketide biosynthesis pathways are complex and involve many enzymes encoded by co-regulated genes clustered at the same locus in the genome (Keller & Hohn, 1997). In the present study, we identified 15 secondary metabolism genes at the avirulence gene ACE1 locus that display the same infection-specific expression pattern as ACE1 (Böhnert et al., 2004). We hypothesized that these 15 genes are involved in the same biosynthetic pathway.
PKSs are responsible for the primary steps of biosynthesis by serial condensation of acyl co-enzyme A. The NRPS module of hybrid PKS-NRPS enzymes catalyzes the attachment of an amino acid to the carboxyl end of the polyketide (Song et al., 2004; Sims et al., 2005). Conserved core motifs of KS, AT, ACP (Mathur & Kolattukudy, 1992) and A, C and PCP (Marahiel et al., 1997) enzymatic domains were found in Ace1 (Böhnert et al., 2004) and Syn2, which are probably functional. Core motifs for the KR and DH domains (Mathur & Kolattukudy, 1992; Aparicio et al., 1996) were less conserved in both enzymes and their functionality cannot be predicted. Ace1 also has an ER domain which is likely to be nonfunctional (EGDDVLGGVG, while the consensus is LIHXXXGGVG; Mathur & Kolattukudy, 1992) and a TE or Claisen-type cyclase domain (GMSGG compared with the consensus motif G(H/Y)SXG; Marahiel et al., 1997; Fujii et al., 2001) which may be involved in the release of the metabolite from Ace1. As Syn2 has no TE domain, release of the metabolite could involve a direct Claisen cyclization of the attached amino acid (Song et al., 2004). Amino acid specificity of bacterial NRPS is correlated with motifs in the adenylation domain, allowing the prediction of the activated amino acid (Stachelhaus et al., 1999). In fungi, only limited data are available to address the amino acid specificity of fungal NRPS modules. Motifs for glycine, serine/alanine and ornithin specificity were described in the adenylation domain of NRPS responsible for siderophore synthesis (Schwecke et al., 2006). Recently, a motif for tyrosine specificity was proposed in the adenylation domain of the PKS-NRPS tenS responsible for the biosynthesis of tenellin (Halo et al., 2008). As these motifs were not found in the sequences of Ace1 and Syn2 NRPS modules, the amino acid they activate cannot yet be predicted. We have compared Ace1 and Syn2 protein sequences to known PKS-NRPSs and constructed a phylogenetic tree of these enzymes (Collemare et al., 2008). Some fungal PKS-NRPS enzymes are involved in the biosynthesis of known polyketides such as lovastatin (LNKS; Kennedy et al., 1999; Hutchinson et al., 2000), fusarin C (FusS; Song et al., 2004), equisetin (EqiS; Sims et al., 2005), aspyridone A (ApdA; Bergmann et al., 2007), pseurotin (PsoA; Maiya et al., 2007) and tenellin (TenS; Eley et al., 2007). These six PKS-NRPSs belong to different PKS-NRPS phylogenetic clades from those of Ace1 and Syn2 (Collemare et al., 2008). Therefore, the metabolites synthesized by Ace1 and Syn2 probably differ from these known molecules.
The ACE1 cluster contains genes that encode proteins with diverse enzymatic activities, suggesting that the metabolite biosynthetic pathway involves many steps. Interestingly, the ACE1 cluster comprises two enoyl reductase-encoding genes, RAP1 and RAP2, that are highly similar to lovC (Table 1), which is an enoyl reductase involved in the biosynthesis of lovastatin in A. terreus (Kennedy et al., 1999). LovC and LNKS interact to produce the nonaketide di-hydroxymonacolin L (Kennedy et al., 1999). Without LovC, LNKS only produces a monomethylated polyunsaturated heptaketide (Auclair et al., 2001). A similar result was obtained when the PKS-NRPS tenS was expressed in Aspergillus oryzae with or without the enoyl reductase encoded by a gene of the tenellin cluster (Halo et al., 2008). Similarly, Ace1 and Syn2 might interact with the enoyl reductases, Rap1 and Rap2, to produce reduced polyketides. Identifying these intermediates will be of great interest to determine how these enoyl reductases modulate the activity of Ace1 and Syn2. The four monoxygenases in the ACE1 cluster belong to two clans of cytochrome P450 (CYP1 and CYP4 to clan CYP605; CYP2 and CYP3 to clan CYP603) specific to M. grisea (Deng et al., 2007). Other modifications of Ace1 and/or Syn2 polyketides may involve their hydroxyl groups as substrates for the putative oxidases Oxr1 and Oxr2, and the methyl transferase Ome1.
The fact that the ACE1 cluster contains two PKS-NRPS-encoding genes is unusual. Only four other polyketide biosynthetic pathways are known to involve two PKSs or one PKS and a FAS. Two PKSs are required for the biosynthesis of T-toxin (Baker et al., 2006) and zearalenone (Kim et al., 2005), but the role of each PKS is unclear. Kim et al. (2005) hypothesize that the first PKS is probably involved in the synthesis of a polyketide used as a starter unit by the second PKS. In the aflatoxin biosynthesis pathway, the starter unit for PksA is hexanoic acid, synthesized by the FAS encoded by two genes in the cluster (Brown et al., 1996). Alternatively, in A. terreus, LNKS and LDKS synthesize two different polyketides, monacolin J and 2-methylbutyrate, respectively, which are fused by a transesterase encoded by lovD (Hutchinson et al., 2000). It is tempting to speculate that the two polyketides produced by Ace1 and Syn2 are fused into a single molecule, as described for lovastatin. Because we did not detect genes encoding enzymes such as LovD in the ACE1 cluster, one polyketide synthesized by either Ace1 or Syn2 could be used as a starter unit by the other PKS-NRPS, as hypothesized for T-toxin and zearalenone (Kim et al., 2005; Baker et al., 2006). An alternative hypothesis is that Ace1 and Syn2 synthesize two different metabolites.
Regulation of the ACE1 cluster
The M. grisea ACE1 locus is a secondary metabolism cluster that contains 15 genes spanning 71 kb. These genes display the same expression pattern, specific to an early stage of infection, which is probably connected to the initiation of the penetration peg formation in appressoria (Fudal et al., 2007). Secondary metabolism clusters are usually regulated by a specific transcriptional regulator from the Zn(II)2Cys6 binuclear zinc finger family encoded by a gene located within such clusters. The ACE1 cluster contains BC2 which encodes a putative transcription factor from the Zn(II)2Cys6 binuclear zinc finger family. Because our attempts to disrupt or silence BC2 were unsuccessful (data not shown), we were not able to determine its role in the regulation of the ACE1 cluster. However, its appressorium-specific expression suggests that it is controlled by a regulatory network, specific to mature appressoria.
The most striking feature of the genomic region carrying the ACE1 cluster is its reduced ability to undergo targeted gene replacement, as observed for ACE1 and SYN2 (1.3 and 4%, respectively). Gene replacement of RAP1, RAP2 and the entire cluster (data not shown) remained unsuccessful despite repeated attempts. Surprisingly, during these experiments, 1 to 4% of the transformants had integrated the gene replacement cassette at one side of the target gene, preserving its functionality (Table 2). This type of integration requires circularization of the linear deletion cassette and its subsequent integration at the target locus through a single crossing-over. As these complex integration events were the only integration events observed at these loci, it is tempting to speculate that double crossing-over cannot occur, which is reminiscent of chromosomal interference observed during meiosis in yeast and ascomycetes (Foss et al., 1993; Roeder, 1995). Therefore, we used a disruption strategy with a circular plasmid that was expected to circumvent the problem. Unfortunately, disruption of RAP1 and RAP2 failed (Table 2), suggesting that the targeted gene replacement rate is even lower with such constructs. Using the Guy11-Δku80 strain, which displays a high targeted gene replacement frequency (up to 98%; Villalba et al., 2008), replacement of ACE1 and RAP1 remained unsuccessful (Table 2). A single integration event was observed for RAP1 without replacement (5.5%), while two rap2::hph disrupted mutants were obtained (5%). As the ACE1 cluster is not located close to the chromosome centromere or telomeres, these results suggest that the ACE1 locus has a reduced ability to undergo homologous recombination, which cannot be rescued by KU80 inactivation (Villalba et al., 2008). It is known in fungi that the frequency of targeted gene replacement varies according to locus (Weld et al., 2006). It is proposed that these variations are related to locus-specific properties, including chromatin status (Moore & Krebs, 2004). In Aspergillus spp., transcriptional regulation of secondary metabolism clusters is under locus-dependent epigenetic control involving the predicted histone methyltransferase LaeA (Bok et al., 2006). Telomere-proximal secondary metabolism clusters are also under the control of the histone deacetylase HdaC (Shwab et al., 2007). This epigenetic regulation of secondary metabolism gene clusters could be an efficient way to synthesize secondary metabolites only when their production is required (Shwab et al., 2007). Therefore, the reduced frequency of targeted gene replacement at the ACE1 locus, regardless of the presence of the KU80 gene, may be a result of locus-specific chromatin modifications that could be involved in the regulation of ACE1 cluster expression.
Table 2. Rate of homologous recombination at the ACE1 locus in Guy11 and Guy11-Δku80 strains during gene replacement and gene disruption strategies
Gene disruption was carried out using a circular vector. The same experiment with a linearized vector did not result in disrupted mutants.
nd, not determined.
< 1 (0/55)
Is the ACE1 cluster an avirulence gene cluster?
Infection-specific expression of the ACE1 cluster suggests that the polyketide produced by this gene cluster has a role in the infection process of M. grisea. However, the Guy11-Δace1 and Guy11-Δsyn2 mutants are as aggressive on susceptible cultivars as Guy11 wild type. These apparently contradictory results might reflect the existence of a functional redundancy between secondary metabolites produced by M. grisea at an early stage of infection. Indeed, M. grisea probably produces other secondary metabolites at this stage of infection, as other PKS-NRPS and PKS genes are expressed specifically at this stage of infection (Collemare et al., 2008; J. Collemare, unpublished). The Guy11-Δsyn2 mutants are still avirulent, demonstrating that this gene is not essential for Pi33-mediated rice recognition. Similarly, CYP2 (cytochrome P450) and MFS1 (transporter) from the avirulent strain 70-15 are probably nonfunctional and may not be required for Pi33 avirulence. The absence of phenotype for rap2::hph disrupted mutants could be a result of a functional redundancy with Rap1 enoyl reductase. We were not able to test this hypothesis as the construction of double mutants was unsuccessful. Therefore, to date only ACE1 has been found to be required for avirulence in Pi33 resistant rice cultivars. These results suggest that the metabolite produced by Ace1 is sufficient to trigger avirulence during the infection of Pi33 resistant rice cultivars. The metabolite produced by Syn2 is probably different from the Ace1 metabolite and is not recognized by Pi33 resistant rice cultivars. As we previously hypothesized that the metabolites produced by Ace1 and Syn2 are fused into a single metabolite, our results suggest that the Syn2 part of this final metabolite is not involved in avirulence, the Ace1 metabolite alone being sufficient for recognition by the host plant.
JC was supported by a MENRT PhD fellowship from the French Ministère de la Recherche, and the work was supported by CNRS and Bayer CropScience. The authors wish to acknowledge Dr Didier Tharreau (UMR BGPI CIRAD-BIOS, Montpellier) for rice infection assays and G. de Gesualdis (Bayer CropScience, Lyon) for rice and barley cultivation. We also acknowledge the referees for their helpful comments on the manuscript.