Candidate effector gene identification in the ascomycete fungal phytopathogen Venturia inaequalis by expressed sequence tag analysis

SUMMARY

The hemi‐biotrophic fungus Venturia inaequalis infects members of the Maloideae, causing the economically important apple disease, scab. The plant–pathogen interaction of Malus and V. inaequalis follows the gene‐for‐gene model. cDNA libraries were constructed, and bioinformatic analysis of the resulting expressed sequence tags (ESTs) was used to characterize potential effector genes. Effectors are small proteins, secreted in planta, that are assumed to facilitate infection. Therefore, a cDNA library was constructed from a compatible interaction. To distinguish pathogen from plant sequences, the library was probed with genomic DNA from V. inaequalis to enrich for pathogen genes, and cDNA libraries were constructed from in vitro‐grown material. A suppression subtractive hybridization library enriched for cellophane‐induced genes was included, as growth on cellophane may mimic that in planta, with the differentiation of structures resembling those formed during plant colonization. Clustering of ESTs from the in planta and in vitro libraries indicated a fungal origin of the resulting non‐redundant sequence. A total of 937 ESTs was classified as putatively fungal, which could be assembled into 633 non‐redundant sequences. Sixteen new candidate effector genes were identified from V. inaequalis based on features common to characterized effector genes from filamentous fungi, i.e. they encode a small, novel, cysteine‐rich protein, with a putative signal peptide. Three of the 16 candidates, in particular, conformed to most of the protein structural characteristics expected of fungal effectors and showed significant levels of transcriptional up‐regulation during in planta growth. In addition to candidate effector genes, this collection of ESTs represents a valuable genomic resource for V. inaequalis.

INTRODUCTION

Venturia inaequalis Cooke (Wint.) is a hemi‐biotrophic fungus that causes disease in members of the Maloideae (Biggs, 1990), the most economically important of which is the apple disease scab or black spot (MacHardy, 1996). The genetic basis of the plant–pathogen interaction between Malus and V. inaequalis follows the gene‐for‐gene model: eight races of the pathogen have been defined by incompatibility, determined by avirulence genes (avr genes), on corresponding host cultivars, each carrying a major resistance gene (R gene), in most cases derived from crab apple varieties (Bénaouf and Parisi, 2000; Bus et al., 2005a,b; Lespinasse, 1994; Parisi and Lespinasse, 1996; Parisi et al., 1993; Roberts and Crute, 1994). Thus, in common with other plant pathogens, this suggests that V. inaequalis produces effector proteins, some of which have been commandeered by the plant for pathogen recognition purposes and initiation of defence.

Effector proteins facilitate infection, often by the suppression of host defence (Chisholm et al., 2006; Jones and Dangl, 2006). In bacterial and oomycete plant–pathogen systems, effector‐mediated suppression of defence has been demonstrated (Bos et al., 2006; DebRoy et al., 2004; Dou et al., 2008; Fu et al., 2007; Nomura et al., 2006). In the majority of filamentous fungal systems, suppression of host resistance has yet to be proven. However, the barley powdery mildew effectors AVR_a10 and AVR_K1 may operate as suppressors as, in susceptible barley cultivars, they increase the number of successful infection sites (Ridout et al., 2006). Moreover, a small secreted protein from the anthracnose fungus Colletotrichum gloeosporioides, which is essential for pathogenicity on Stylosanthes, appears to act during the initial biotrophic phase of infection to suppress host cell death (Stephenson et al., 2000). In addition, the Cladosporium fulvum effector Avr4 is required for full virulence on tomato, and is believed to be involved in counter‐defence, protecting against host chitinases (van Esse et al., 2007). Houterman et al. (2008) recently reported the suppression of R‐gene‐based resistance by an effector from Fusarium oxysporum f. sp. lycopersici. Effector genes, including those with a concomitant avirulence function, have yet to be cloned from V. inaequalis.

During infection, V. inaequalis penetrates the cuticle and forms infection hyphae that ramify above the epidermal cells, from which stromata form. Stromata are pseudo‐parenchymatous structures made up of laterally dividing cells, markedly different to the tubular hyphae found in vitro (Kucheryava et al., 2008). Stromata are presumed to be required in order to attain nutrients from the subcuticular space, and give rise to conidia that are disseminated through the ruptured cuticle to further the infection cycle. As no penetration of plant cells occurs, effectors are presumably secreted from the hyphae and stromata into the subcuticular space, where they may exert their effect extracellularly, or may be taken up by the plant cell by an unknown mechanism.

In the V. inaequalis–apple pathosystem, it is hypothesized that the plant has evolved to recognize a subset of these effectors (avr gene products) either directly or indirectly by R gene products, which trigger a hypersensitive response (HR). It is possible that recognition may take place extracellularly, as in the C. fulvum–tomato interaction (Kruijt et al., 2005). Alternatively, interaction of effector and R gene product may occur intracellularly, as secreted fungal proteins have been shown to be taken up by plant cells in some pathosystems (Catanzariti et al., 2006; Dodds et al., 2004; Kemen et al., 2005; Ridout et al., 2006). The former scenario is supported by the finding that cloned apple R genes (forming a cluster at the Vf locus), which confer resistance to V. inaequalis (Vinatzer et al., 2001; Xu and Korban, 2002), all belong to the same class of R gene as the Cf genes from tomato (Dangl and Jones, 2001). Given this similarity and the parallels in parasitic strategy adopted by each pathogen, an extracellular site of action seems most plausible.

The isolation of proteinaceous effectors from the subcuticular space has proved problematic (Fitzgerald, 2004; Win, 2003). Gau et al. (2004) were able to isolate apple pathogenesis‐related (PR), but not V. inaequalis, proteins from the apoplastic fluid of infected leaves. A map‐based cloning strategy to clone effectors has been adopted by Broggini et al. (2007), and has resulted in the isolation of a 330‐kb contig comprising 12 bacterial artificial chromosomes containing the VirQ5 marker for the avr gene AvrRvi1. However, the avr gene itself has yet to be identified.

An alternative approach for the identification of effector candidate genes involves the construction of expressed sequence tag (EST) libraries followed by bioinformatic analysis. This approach has proven to be valuable in the study of organisms where very little sequence data are available for either host or pathogen (Cramer et al., 2006).

The identification of effector genes by bioinformatic analysis of ESTs can be facilitated by the presence of conserved motifs common to effectors. For example, the RxLR motif of oomycete pathogens is involved in the delivery of effectors to the host cell (Bhattacharjee et al., 2006; Jiang et al., 2008; Rehmany et al., 2005; Vleeshouwers et al., 2008; Whisson et al., 2007), and can be used to identify novel candidate effectors (Torto‐Alalibo et al., 2007). In contrast with the oomycetes, filamentous fungal effector proteins identified to date share little homology with one another, or with sequences present in the public domain, although they do share some common characteristics. They are usually small: less than 400 amino acids in length when mature. They may be cysteine rich, which may aid protein stability in the extracellular environment by disulphide bond formation (Rep, 2005; Templeton et al., 1994). They typically possess an N‐terminal signal peptide for secretion, although there are exceptions, including AVR_a10 and AVR_K1 from Blumeria graminis f. sp. hordei (Ridout et al., 2006).

In this study, cDNA libraries were constructed from infected leaf material and from in vitro‐grown V. inaequalis, the latter to aid in origin assignment for sequences from the libraries derived from infected leaf material. The in vitro‐grown libraries include a suppression subtractive hybridization (SSH; Diatchenko et al., 1996) library enriched for cellophane‐induced genes, as V. inaequalis produces structures in cellophane morphologically similar to those found in planta, suggesting that growth in cellophane may be an in vitro model of infection by V. inaequalis (Kucheryava et al., 2008). Indeed, two cellophane‐induced genes have been shown to be up‐regulated in planta (Kucheryava et al., 2008). The SSH library was included to circumvent the common problem of a scarcity of pathogen sequences being identified when sequencing ESTs from infected plant material (Cramer et al., 2006). An SSH strategy has previously been proven to be invaluable for the study of fungal–plant interactions (van den Berg et al., 2007; Cramer et al., 2006; Yakovlev et al., 2008). The dataset reported herein comprises 2090 tentative non‐redundant sequences following clustering from 4215 ESTs. It represents the first major analysis of ESTs from the V. inaequalis–apple interaction, which identifies a profile of genes expressed by the pathogen during pathogenesis, including 16 candidate effector genes that have been identified on the basis of their putative small size (< 400 amino acids in length), secretion (based on the SignalP algorithm; Bendtsen et al., 2004; Nielsen et al., 1997) and the presence of cysteine residues.

RESULTS

cDNA libraries

Five distinct cDNA libraries were created and their sequences entered into the Plant and Food Research fungal sequence database. Two libraries were prepared from infected leaves: ABEA and ABEB (in planta libraries). ABEB was prepared from the ABEA library by probing the ABEA library with genomic DNA from V. inaequalis to enrich for fungal sequences. Two further libraries [IAAA and MAAB; potato dextrose broth (PDB) in vitro libraries] were prepared from in vitro‐grown mycelium from PDB shake cultures. The final library (MAAD; SSH cellophane in vitro library) comprised ESTs from the SSH library enriched for cellophane‐induced genes. The SSH method (Diatchenko et al., 1996) was utilized to enrich for cellophane‐induced genes, the profile of which would potentially mirror that observed during growth in planta. All libraries were non‐directional and sequenced from only one direction (Table 1).

Table 1.

Summary of the Plant and Food Research Venturia inaequalis expressed sequence tag (EST) libraries.

Library	Description	Total # fungal* ESTs	Min. EST length	Max. EST length	Average sequence length	# nr‡	# contigs (#ESTs)	# singletons	Redundancy§
ABEA	7‐week‐old Malus × domestica (‘Royal Gala’) Vi race 1 (MNH 135), 17 days post‐inoculation (dpi)	587 (2342)†	145	1835	572	432	55 (210)	377	26%
ABEB	ABEA probed with race 1 (MNH 135) Vi gDNA to select fungal sequences	350 (404)	197	1472	558	227	74 (197)	153	35%
IAAA	Vi race 1 (MNH 120) in vitro, 30 dpi	1311	69	755	596	625	322 (1008)	303	52%
MAAB	Vi race 1 (MNH 120) in vitro, 30 dpi	861	102	1268	635	689	99 (271)	590	20%
MAAD	Vi race 1 (MNH 120) suppression subtractive hybridization library enriched for cellophane‐induced genes, 28 dpi	1106	128	1130	544	582	201 (725)	381	47%

^* This value is putative for the in planta libraries, as designations of origin are based on bioinformatic analysis or clustering with in vitro library sequences, with the remainder of the sequences having no known similarities and therefore no firm assignation as fungal or plant.

^† Number in parentheses denotes total number of sequences in the library (both plant and fungal).

^‡ nr, non‐redundant sequence, with respect to individual libraries.

^§ Redundancy: total number of ESTs – (total number of contigs + total number of singletons)/total number of sequences (Cramer et al., 2006).

A total of 4215 ESTs was identified from the five libraries and, following clustering using the cap3 program (Huang and Madan, 1999), were resolved into 2090 non‐redundant sequences (Table 1). When individual libraries were analysed, the 587 putatively fungal ESTs from ABEA were resolved into 432 non‐redundant sequences, including 55 contigs and 377 singletons. The 350 ESTs from ABEB were assembled into 227 non‐redundant sequences, comprising 74 contigs and 153 singletons. Assembly of the 1311 ESTs from IAAA into 625 non‐redundant sequences revealed 322 contigs and 303 singletons, and the 861 ESTs from MAAB were assembled into 689 non‐redundant sequences, including 99 contigs and 590 singletons. The 1106 ESTs from MAAD were assembled into 582 non‐redundant sequences with 201 contigs and 381 singletons (Table 1).

A redundancy level for each library was calculated, with the lowest rate (20%) exhibited by the MAAB in vitro library, suggesting that further sequencing of clones from this library will reveal additional novel sequences. The rates for the two in planta libraries were 26% and 35% for ABEA and ABEB, respectively. The usefulness of further sequencing of the ABEA library to identify additional novel fungal genes, even though the level of redundancy is 26%, is questionable, given that only 25% of the ESTs sequenced were putatively fungal in origin. However, it is a valuable resource of plant genes, some of which may be specifically expressed in the V. inaequalis–apple interaction. Further sequencing of ABEB would be more expedient, with 87% fungal genes, notwithstanding the higher level of redundancy, which at 35% still intimates that additional novel sequences important to the fungus during host colonization could be identified. The rates for the IAAA PDB and SSH cellophane in vitro libraries were 52% and 47%, respectively.

Origin of ESTs from infected plant material‐derived cDNA libraries

Of the 2746 ESTs in the two infected leaf libraries, only 598 (340 non‐redundant sequences) were categorized as fungal with a significant level of confidence. This was accomplished by either clustering with V. inaequalis in vitro library ESTs (Fig. 1) or by bioinformatic analysis.

Figure 1.

The distribution of putatively fungal expressed sequence tags (ESTs) and non‐redundant sequences (nr) between the in planta libraries (ABEA and ABEB), the potato dextrose broth (PDB) in vitro libraries (IAAA and MAAB) and the suppression subtractive hybridization (SSH) cellophane in vitro library (MAAD). Where there are nr with representative ESTs from more than one library (overlapping sectors), the numbers given are totals from all contributing libraries.

Three hundred ESTs, which formed 153 non‐redundant sequences that included members from the in vitro libraries, were recorded as fungal (Fig. 1). blastp (Basic Local Alignment Search Tool) searches (Altschul et al., 1997) of the predicted translations of the ESTs against the National Center for Biotechnology Information (NCBI) RefSeq (Pruitt et al., 2007) and UniProt UniRef90 (Suzek et al., 2007) databases revealed a further 298 ESTs as most likely to be fungal in origin, which, following clustering, were resolved into 187 non‐redundant sequences. Although a relaxed stringency level (≤ 1e‐5) was used to demarcate ESTs with similarities to fungal sequences in the public domain, the majority of sequences were significantly more similar, with E‐values of ≤ 1e‐10; seven sequences in ABEA and four in ABEB recorded E‐values in the range ≥ 1e‐10 to ≤ 1e‐5.

A further 339 ESTs that could be assembled into 293 non‐redundant sequences from the two in planta libraries showed no significant similarity to any known sequences, including apple sequences. Although the fungal origin of these sequences cannot be confirmed, they represent novel sequences and, as such, are a putative source of effectors, as those from filamentous fungi are often unique. They were included in the putatively fungal dataset to preclude the loss of potential candidate effector genes. Therefore, the total number of ESTs that were classified as putatively fungal was increased to 937, which could be assembled into 633 non‐redundant sequences.

Characterization of the libraries

Of the 2172 ESTs in the PDB in vitro libraries (IAAA and MAAB), 26%, comprising 369 non‐redundant sequences, showed no similarity to sequences in the public domain using a threshold E‐value of ≤ 1e‐5. In the cellophane in vitro library (MAAD), this level was 22%, comprising 132 non‐redundant sequences.

The majority of ESTs (70%) present in the IAAA and MAAB libraries were unique to those libraries. Similarly, 77% of ESTs in the MAAD library and 68% of ESTs from the in planta libraries were also unique (Fig. 1). As the cellophane in vitro library was constructed using SSH methodology, enriching for those genes expressed during morphological differentiation on cellophane and subtracting those genes common to both V. inaequalis growing on cellophane and in PDB, the fact that relatively few, only 94, non‐redundant sequences were constructed from ESTs belonging to both sets of libraries is expected, indicating efficient subtraction. Indeed, the efficiency of subtraction was such that the proportion of ESTs identified as cell function genes in the MAAD library was far less than that in the PDB in vitro libraries (Table 2). However, given the hypothesis that growth in cellophane is a model of infection, the fact that only 43 (7%) non‐redundant sequences had EST members from both the in planta and cellophane in vitro libraries was unexpected. This may reflect the uniqueness of the gene profile required for pathogenicity and that growth in cellophane does not faithfully replicate all aspects of growth in planta. However, the MAAD library was constructed from rinsed cellophane sheets, the rinsing step resulting in the removal of surface hyphae, conidiophores and conidia that would be present in the infected leaf material used for in planta library construction. Consequently, the MAAD library may have a lower representation of sequences associated with the structures removed by rinsing. Alternatively, this finding may be a result of the isolates used to construct the two types of library, even with the threshold for non‐redundant sequence alignment set at 95% identity. Non‐isogenic race 1 isolates were used: MNH135 was used for in planta library construction, whereas MNH120 was used for in vitro library construction. Given that the two isolates used are both categorized as race 1 isolates, they should still share common effectors with respect to avirulence determinants that are recognized by cognate major R genes, and it is unlikely that these effectors would show significant sequence divergence.

Table 2.

Occurrence of cell function genes in the potato dextrose broth (PDB) in vitro libraries (IAAA and MAAB) compared with that in the suppression subtractive hybridization (SSH) cellophane in vitro library (MAAD), as an indication of subtraction efficiency in the MAAD library. Quantification was based on a keyword search using ‘actin’, ‘ribosomal protein’, ‘glyceraldehyde 3‐phosphate dehydrogenase’ (gapdh), ‘ubiquitin’ and ‘ubiquitin conjugating enzymes’ versus the annotations assigned to each expressed sequence tag (EST) based on a blastp algorithm (Altschul et al., 1997) search of the National Center for Biotechnology Information (NCBI) RefSeq (http://blast.ncbi.nlm.nih.gov/Blast.cgi; Pruitt et al., 2007) and UniProt UniRef90 (http://www.ebi.ac.uk/uniprot/; Suzek et al., 2007) databases.

Cell function gene	Occurrence of keyword in each library
	IAAA	%	MAAB	%	MAAD	%
Actin	2	0.15	2	0.23	0	0
Ribosomal protein	78	5.9	66	7.69	7	0.63
gapdh	1	0.07	6	0.7	0	0
Ubiquitin	10	0.76	4	0.47	1	0.09
Ubiquitin‐conjugating enzymes	2	0.15	1	0.12	0	0
Total # ESTs	1311		861		1106

Bioinformatic analysis of candidate effector genes

The strategy adopted for the selection of candidate effector genes is outlined in Fig. 2. Of 633 non‐redundant sequences from the cDNA libraries that were classified as either fungal or of unknown origin, 96 were selected on the basis of possessing a putative signal peptide following analysis by the SignalP algorithm (Bendtsen et al., 2004; Nielsen et al., 1997). From these, 21 putative effector candidates were selected on the basis of being novel [having no similar sequences in either the NCBI RefSeq (Pruitt et al., 2007) or UniProt UniRef90 (Suzek et al., 2007) database] and, where possible, possessing cysteines and being small (less than 400 amino acids), although candidates were not required to meet all of these criteria. Indeed, several required further sequencing to enable the identification of full‐length putative open reading frames [ORFs; utilizing the Kozak sequence (Balance, 1986; Kozak, 1986) surrounding the 5′ ATG of the longest putative ORF to confirm start codons], and so could not be classified initially according to the size of the encoded product. These sequences were then subjected to a further round of blastp similarity searches. This revealed that two were probably derived from apple, as full‐length ORF sequences revealed strong similarities to Arabidopsis thaliana proteins. A further three candidates had been identified in previous studies (Kucheryava et al., 2008; Win, 2003; Win et al., 2003), and therefore these five ESTs were discounted from further analysis. The remaining 16 candidate effectors were given identification numbers (Venturia inaequalis candidate effector: Vice1–16) and are listed in Table 3.

Figure 2.

Strategy used for the identification of candidate effectors from Venturia inaequalis expressed sequence tags (ESTs). ORF, open reading frame.

Table 3.

Summary of the attributes of selected Venturia inaequalis candidate effector genes from the cDNA libraries.

ID number* (# ESTs in nr)	Library of origin of nr members	Best blastp match† (annotation, organism, accession number)	E‐ value‡	Length of ORF§ (amino acids)	Length of signal peptide¶	# Cysteine residues**	GPI anchor site (position)††
Vice1 (1)	1 × ABEA	No significant match‡		103	20/20	5	–
Vice2 (1)	1 × ABEA	Hypothetical protein, Aspergillus oryzae RIB40, XP_001821705.1	2e‐27	210	16/16	NA‡‡	NA
Vice3 (1)	1 × ABEA	Hypothetical protein SNOG_03711, Phaeosphaeria nodorum SN15, XP_001794262.1	6e‐31	403	19/22	2	–
Vice4 (1)	1 × ABEA	No significant match		88	19/19	8	–
Vice5 (1)	1 × ABEA	Hypothetical protein SS1G_02025, Sclerotinia sclerotiorum 1980, XP_001597829.1	8e‐38	245	19/19	6	–
	1 × MAAB
	1 × MAAD
Vice6 (1)	1 × ABEB	Conserved hypothetical protein, Pyrenophora tritici‐repentis Pt‐1C‐BFP, XP_001931281.1	7e‐11	174	17/17	0	+ (149)
Vice7 (5)	1 × ABEB	No significant match		91	16/17	6	–
	3 × IAAA
	1 × MAAB
Vice8 (1)	1 × ABEB	No significant match		126	18/18	9	–
Vice9 (1)	1 × ABEB	No significant match		261	15/15	3	+ (240)
Vice10 (7)	5 × ABEB	Hypothetical protein, Yarrowia lipolytica, XP_504084.1	1e‐18	297	17/17	4	–
	1 × MAAB
	1 × MAAD
Vice11 (13)	1 × MAAB	No significant match		260	20/20	8	+ (233)
	12 × MAAD
Vice12 (1)	1 × MAAD	No significant match		61	16/17	2	–
Vice13 (1)	1 × MAAD	No significant match		68	29/29	NA	NA
Vice14 (10)	1 × ABEA	Conserved hypothetical protein, Aspergillus terreus, XP_00121059.1	1e‐42	334	15/15§§	6	–
	2 × ABEB
	4 × IAAA
	3 × MAAD
Vice15 (2)	2 × MAAD	No significant match		51	21/21	NA	NA
Vice16 (9)	9 × MAAD	Hypothetical protein PTRG 01937, Pyrenophora tritici‐repentis, XP_001932270.1	3e‐63	211	15/15	4	–

^* EST, expressed sequence tag; nr, non‐redundant sequence; Vice, Venturia inaequalis candidate effector.

^† blastp searches were carried out against the National Center for Biotechnology Information (NCBI) RefSeq (http://blast.ncbi.nlm.nih.gov/Blast.cgi; Pruitt et al., 2007) and UniProt UniRef90 (http://www.ebi.ac.uk/uniprot/; Suzek et al., 2007) databases.

^‡ E‐values were recorded (matches classed as significant) if they were ≤ e‐05.

^§ The length of each peptide encoded by the predicted open reading frame (ORF) was recorded, including the predicted signal peptide.

^¶ The length of the putative signal peptide is given based on the neural networks (NN) and hidden Markov models (HMM) of the SignalP server (http://www.cbs.dtu.dk/services/SignalP/; Bendtsen et al., 2004; Nielsen et al., 1997).

^** The number of cysteine residues in the putative protein encoded by the ORF.

^†† The presence of putative C‐terminal GPI (glycosylphosphatidylinositol) lipid modification (anchor) sites was analysed using the big‐PI fungal server (http://mendel.imp.ac.at/gpi/fungi_server.html; Eisenhaber et al., 2004) with a threshold score of ≥ 6: +, site identified with location in parentheses; –, no site identified.

^‡‡ NA, not analysed because of predicted transmembrane domains.

^§§ The TargetP server predicted that this EST may be localized in the mitochondrion and not directed to the secretory pathway.

Seven candidates that were originally selected as novel were found to have significant similarity to other fungal proteins following the additional sequencing and similarity searches. However, the best matches were all against hypothetical proteins without positive annotations, and so these candidates were retained for further analysis. In addition, Vice2 revealed significant blastp (2e‐27) similarity to plastocyanin‐like proteins of other fungal species. However, no type 1 copper‐binding motif was found within the predicted protein sequence of Vice2 using the type 1 copper‐binding signature (Pattern PS00196: [GA]‐x(0,2)‐[YSA]‐x(0,1)‐[VFY]‐{SEDT}‐[C]‐x(1,2)‐[PG]‐x(0,1)‐H‐x(2,4)‐[MQ]) as retrieved from PROSITE (http://expasy.org/cgi‐bin/prosite/). Vice3 was found to be similar (blastp E‐value of 6e‐31) to acetylornithine deacetylase, succinyl‐diaminopimelate desuccinylase and related deacylases. The active site residues of experimentally confirmed acetylornithine deacetylase and succinyl‐diaminopimelate desuccinylase proteins from the bacterium Escherichia coli were found to align with identical residues in Vice3. However, the consensus pattern of active site residues also aligned with other peptidases, such as carboxypeptidase G2 and peptidase T, which may suggest that this protein has some form of peptidase activity. The protein sequence encoded by Vice5 demonstrated significant similarity (8e‐38) to a cysteine‐rich hypothetical protein of Sclerotinia sclerotiorum. All six C‐terminal cysteine residues of Vice5 were strictly conserved with this hypothetical protein. Conservation of cysteine residue spacing was also observed between Vice10 and similar protein sequences predicted to possess a rare lipoprotein A (RlpA)‐like double‐psi β‐barrel domain, the function of which is unknown.

The majority of the results generated by the SignalP algorithm were corroborated by the TargetP server with candidate effectors predicted to enter the secretory pathway, with the exception of Vice14 (Table 3). In this case, a putative mitochondrial subcellular localization may be possible, although the lowest reliability class was recorded for the result, with similar scores recorded for a mitochondrial (0.315) and a secretory pathway (0.277) location.

Analysis using the TMHMM server revealed the presence of transmembrane domains in Vice2, Vice13 and Vice15, implying that they are not secreted. These three candidate effectors were therefore discounted from further analysis.

All candidate effector proteins contained cysteine residues, except Vice6. The majority of the candidates that possessed two or more cysteine residues were predicted to possess a disulphide bond by the DISULFIND server, with the exception of Vice3, Vice9, Vice12 and Vice14. Putative disulphide bond spacing was conserved between similar proteins in the public domain from Phaeosphaeria nodorum and Vice5 and Vice10.

C‐terminal glycosylphosphatidylinositol (GPI) lipid modification (anchor) sites were predicted for the protein sequences Vice6, Vice9 and Vice11 using the big‐PI server (Table 3). InterProScan searches revealed motifs in six of the candidate effectors (Table 4). Several of the servers contributing to the InterProScan search revealed that Vice3 comprises domains with similarities to metallopeptidases and peptidase dimerization motifs. Similarities to acetylornithine deacetylase were also identified, which was consistent with the blastp results. Vice11 showed similarities to motifs associated with C‐type lectins, and Vice14 showed similarities to domains associated with allergen V5 (vespid wasp allergen)/Tpx‐1 (mammalian testis‐specific protein), a plant PR protein and sterol carrier protein (SCP)‐like extracellular protein.

Table 4.

InterProScan results (http://www.ebi.ac.uk/Tools/InterProScan/; Mulder et al., 2007; Quevillon et al., 2005) for candidate effectors from Venturia inaequalis. Gene ontology (GO) classifications are shown, where known, indicating cellular component, biological process and molecular function applicable to motifs identified by the InterPro member databases.

ID number	InterPro member database and search method*	Motif	GO cellular component	GO biological process	GO molecular function	Amino acid residues†	E‐value
Vice3	HMMPfam	PF01546 Peptidase_M20 IPR002933		Proteolysis (GO:0006508)	Metallopeptidase acitivity (GO:0008237)	109–401	7.9e‐17
		PF07687 M20_dimer IPR011650 Peptidase M20, dimerization			Hydrolase activity (GO:0016787), protein dimerization activity (GO:0046983)	210–318	8.8e‐23
	HMMPanther	PTHR11014:SF5 Acetylornithine deacetylase				44–402	1.3e‐79
		PTHR11014 Peptidase M20 family member				44–402	1.3e‐79
	ScanRegExp	PS00758 ARGE_DAPE_CPG2_1 IPR001261 Arg/DapE/ACY1/CPG2/ YscS. Conserved site		Proteolysis (GO:0006508)	Metallopeptidase activity (GO:0008237)	108–117
Vice11	ScanRegExp	PS00615 C_type_lectin_1			Binding (GO:0005488)	36–64
Vice14	HMMSmart	SM00198 SCP IPR001283 Allergen V5/Tpx‐1 related	Extracellular region (GO:0005576)			164–314	3.3e‐21
	HMMPfam	PF00188 SCP IPR014044 SCP‐like extracellular				170‐306	4.2e‐21
	ScanRegExp	PS01009CRISP_1 IPR001283 Allergen V5/Tpx‐1 related	Extracellular region (GO:0005576)			266–276	8e‐5
	FPrintScan	PR00837 V5TPXLIKE IPR001283 Allergen V5/Tpx‐1 related	Extracellular region (GO:0005576)			185–203	4.3e‐11
						265–281	4.3e‐11
						301–314	4.3e‐11
	HMMPanther	PTHR10334:SF6 Plant pathogenesis‐related				174–317	2.1e‐30
		PTHR10334 Cysteine‐rich secretory protein (CRISP/SCP/TPX1)‐related IPR001283 Allergen V5/Tpx‐1 related	Extracellular region (GO:0005576)			174–317	2.1e‐30
	BlastProDom	PD000542 Q7SG93 NEUCR Q7SG93; IPR001283 Allergen V5/Tpx‐1 related	Extracellular region (GO:0005576)			233–320	2e‐11

^* InterPro member databases and search methods: BLASTProDom, a gapped blastp search of protein families; FprintScan, a scan against the fingerprints (groups of motifs) in the PRINTS database; HMMPanther, a scan of the PANTHER protein families; HMMPfam, a scan of the hidden Markov models (HMMs) in the Pfam Protein Families database; HMMSmart, a scan of the HMMs present in the SMART domain/domain families database; ScanRegExp, a scan of the regular expressions in the PROSITE protein families and domains database.

^† Amino acid residues in the candidate effector that match the motif.

The SAM‐T06 server was used to screen for similarities between predicted secondary and tertiary candidate effector structure and solved structures. Consistent with blastp and InterProScan results, the protein sequence of Vice3 demonstrated moderate structural similarity to peptidase M20A proteins (4.1e‐46). Of interest, Vice10 demonstrated moderate structural similarity (3.4e‐11) to endoglucanases, Barwin (Barley wound‐induced plant defence protein), expansins and pollen allergens. InterProScan results were also corroborated by similarities detected using candidate effector Vice14, as it demonstrated moderate structural similarity to a V5 allergen (1.4e‐25) belonging to the PR‐1‐like superfamily with an SCP‐like extracellular domain.

Genomic DNA extraction and PCR analysis

A fungal origin was confirmed for those candidate effectors (Vice1, Vice3, Vice4, Vice6, Vice8 and Vice9) for which coding sequences were present in the in planta libraries only: amplification of a PCR product from genomic DNA of V. inaequalis isolate MNH120 was recorded (results not shown).

Leaf infection and mRNA expression analysis of candidate effector genes

Leaf samples were collected during a detached leaf infection assay at 5 and 10 days post‐inoculation (dpi) and were examined microscopically. Concomitant expression of each of the candidate effector genes was measured by real‐time quantitative reverse transcriptase‐polymerase chain reaction (qRT‐PCR), and compared with that during growth in PDB at 10 dpi. Differences in expression were deemed significant at P = 0.01. At 5 dpi, spores of V. inaequalis had germinated and penetrated the cuticle, and subcuticular hyphae had ramified in the subcuticular space. Small stromata consisting of a few enlarged cells were beginning to be formed (Fig. 3). At 10 dpi, stromata were largely mature, with conidiophores arising from them bearing conidia (Fig. 3).

Figure 3.

Infection by Venturia inaequalis during a detached leaf assay at (a) 5 days post‐inoculation (dpi), with stromata beginning to form, and (b) 10 dpi, with sporulating stromata. Scale bar, 50 µm.

Candidate effector mRNAs for Vice4, Vice12 and Vice16 were all significantly up‐regulated during in planta growth at both 5 and 10 dpi, whereas Vice5 was only up‐regulated at 5 dpi, albeit with only a 1.6‐fold increase, with no significant difference in expression between in planta and PDB growth at 10 dpi (Fig. 4). The expression of Vice1, Vice3 and Vice7 candidate effector mRNAs during growth in planta was not significantly different from that during growth in PDB at either of the infection time points. The expression of five candidate effector mRNAs (Vice6, Vice9, Vice10, Vice11 and Vice14) was significantly lower during growth in planta than during growth in PDB at both time points analysed, whereas the expression of Vice8 mRNA was significantly reduced at 5 dpi, but not significantly different at 10 dpi. No expression of candidate effector mRNAs was recorded on leaves inoculated with sterile distilled water (SDW; data not shown).

Figure 4.

The average expression of four Venturia inaequalis candidate effector (Vice) genes at two time points during detached leaf infection compared with that during growth in potato dextrose broth (PDB). A gene expression normalization factor was calculated for each sample based on the geometric mean of two stably expressed reference genes (β‐tubulin; EU853839) and 60S ribosomal protein L12 (EU853840) using geNorm software v3.4 (Vandesompele et al., 2002). Averages were calculated from six expression measurements from each time point of infection or growth stage, gathered from two independent infection/growth assays; each expression measurement was the average of three biological samples (each sample comprising three infected leaves or material gathered from three flasks of PDB), and the polymerase chain reaction (PCR) for each sample was conducted in triplicate. The error bars represent the standard deviation of the six expression measurements. The expression of each of these genes during infection was significantly different (P = 0.01) from that during growth in PDB, with the exception of Vice5 (light grey) at 10 dpi, which only showed a 1.39‐fold increase. Vice4, diagonal lines; Vice5, light grey; Vice12, black; Vice16, horizontal lines.

DISCUSSION

This study represents the first large‐scale analysis of ESTs derived from the V. inaequalis–apple pathosystem. It resulted in the identification of 937 putatively fungal ESTs that were resolved into 633 non‐redundant sequences following clustering. An EST approach has proven to be successful in identifying genes involved in other fungal pathosystems (Cramer et al., 2006; Ebbole et al., 2004; Hu et al., 2007; Kleemann et al., 2008; Li et al., 2004). Moreover, it has proven to be successful in the identification of some phytopathogen effector genes (Armstrong et al., 2005; Bos et al., 2003; Catanzariti et al., 2006; Shan et al., 2004). For example, Catanzariti et al. (2006) generated a Melampsora lini haustorium‐specific EST library, as the HR of rust resistance appeared to coincide exclusively with emerging haustoria, to enable the identification and cloning of AvrM, AvrP4 and AvrP123‐A, three effector genes of M. lini (Catanzariti et al., 2006). Armstrong et al. (2005) cloned and characterized the Avr3a effector gene of Phytophthora infestans using a combination of EST database mining (Torto et al., 2003) and an association genetics approach originally developed by Bos et al. (2003).

Despite these successes, the largest drawback of an EST strategy is the difficulty encountered when assigning the origin of sequences according to organism. The most valuable resource in aiding this assignment is the complete genome sequence of the organisms involved in the interaction. Cramer et al. (2006) demonstrated that, in the Alternaria brassicicola–Brassica olearacea interaction, 30% more fungal genes were identified in a library from infected plant material with a whole‐genome draft sequence than using only blast searches of the same library. Although advances have been made in sequencing the apple genome, a complete sequence has yet to be released, and the genome sequencing of V. inaequalis has yet to be initiated.

Various strategies were employed in this study to overcome these limitations. Firstly, the ABEA EST library constructed from infected leaf tissue was probed with genomic DNA from V. inaequalis to enrich for fungal genes expressed in planta to create the ABEB library. A DNA blot approach has proven to be successful for other phytopathogens (Hahn and Mendgen, 1997; Rauyaree et al., 2001; Thara et al., 2003; Zhang et al., 2003). This resulted in a 3.5‐fold increase in the percentage of fungal sequences in the ABEB library compared with the ABEA library, as 87% of ESTs were designated fungal in the former and 25% in the latter. Secondly, libraries were constructed from V. inaequalis growing axenically to aid the designation of fungal sequences. The ESTs generated from PDB‐grown V. inaequalis enabled 824 ESTs from the infected leaf material libraries to be designated as fungal. This approach, however, suffers from the drawback that effector genes may have a tendency to be expressed exclusively in planta. Thus, further refinement of this strategy was the use of the SSH library enriched for cellophane‐induced genes. This approach was adopted as growth in cellophane was thought to be a model of in planta growth, given the differentiation of infection‐like structures during growth in cellophane (Kucheryava et al., 2008). Similarities between sequences in the in planta and SSH cellophane‐enriched libraries enabled the designation of a further 98 ESTs. The SSH library strategy was only partially successful: although it resulted in the designation of additional sequences as fungal from the infected leaf material libraries, this was not to the level expected if growth in cellophane absolutely mimics that in planta. However, it resulted in the identification of seven candidate effector genes that would otherwise not have been found.

Alternative methods for the differentiation of pathogen and host sequences in a mixed population have been used previously. These include the use of the difference in average GC/AT ratios for plant and pathogen genes (Qutob et al., 2000), or codon bias, notably the third ‘wobble’ position of the codon (Emmersen et al., 2007; Friedel et al., 2005; Hsiang and Goodwin, 2003; Li et al., 2004; Maor et al., 2003). Initial assessment of genes from V. inaequalis and apple indicated that there was a significant difference between GC/AT ratios and codon usage profiles in genes from the two species (results not shown). This may be used in future analyses of ESTs from this interaction to differentiate between host and pathogen sequences. However, it is important to note that, although GC content and codon usage bias differ between species, vast differences can exist between different genes of a single species or different regions within a gene (Friedel et al., 2005), leading to false designations.

As it is assumed that candidate effectors will be secreted into the subcuticular space during infection by V. inaequalis, and may subsequently be translocated into the host cytoplasm, the SignalP algorithm was used to scan the putative V. inaequalis sequences identified from the mixed population of ESTs from the in planta and in vitro libraries for N‐terminal signal peptide sequences. Kleemann et al. (2008) outlined the inherent pitfalls in the use of such an in silico approach; accurate identification of the N‐terminal methionine can be problematic as ORF prediction algorithms can be unreliable, especially with fungal sequences. Incorrect ORF prediction can lead to internal transmembrane domains being identified as signal peptides by the SignalP algorithm. Hence, the longest putative ORF with a 5′ ATG with conservation of Kozak sequence specific to filamentous fungi (Balance, 1986; Kozak, 1986) was used to identify the most likely full‐length ORF (data not shown). An alternative approach for the identification of secreted proteins would be to use the ‘signal sequence trap’ methodology (Klein et al., 1996). This yeast‐based genetic screen has been used to confirm the presence of functional secretion signals in two plant pathogens to date: Pseudorobillarda sojae (Lee et al., 2006) and Uromyces fabae (Link and Voegele, 2008).

This approach, focusing as it does on the selection of ESTs with putative signal peptides, could result in some effectors being overlooked. The paradigm for filamentous fungal effectors is that they are small, often cysteine‐rich, secreted proteins. Indeed, the majority of fungal effectors fit this paradigm. There are, however, notable exceptions: AVRa10 and AVRk1 from the barley powdery mildew fungus B. graminis f. sp. hordei. These effectors trigger hypersensitive cell death when expressed in the cytoplasm of resistant barley cultivars carrying the R genes Mla10 and Mlk1. How they are secreted is unknown, as they lack the typical signal peptide signature (Ridout et al., 2006). These effectors do not comply with the paradigm outlined above and would not be identified in the search strategy employed in this study. However, the use of the SignalP algorithm led to the identification of 96 non‐redundant sequences which, following blastp searches, were reduced to 16 candidate effector genes.

Candidate effectors were initially selected on the basis of being novel. However, some of these were subsequently found, on identification of the complete ORF sequence, to have significant similarities to hypothetical fungal genes from other species. This may not preclude these sequences from being effectors, however, as the steady increase in the number of fungal phytopathogen genomes being sequenced increases the likelihood that orthologues will be found in other species. Indeed, the C. fulvum lysin motif effector Ecp6 has orthologues from 11 filamentous fungal species, including seven plant pathogens (Bolton et al., 2008).

Although the most significant matches, when any were recorded, for each candidate effector protein (as revealed by blastp searches) were to a hypothetical protein from other fungi, several proteins with less significant E‐values were identified, and these possessed annotated domains. For example, the full‐length protein sequence of Vice3 demonstrated significant similarity to proteins of the peptidase subfamily M20A involved in amino acid metabolism. Although, in principle, any foreign molecule could act as a potential effector, Vice3 does not lend itself to an obvious role in virulence or pathogenicity if these similarities represent a true conservation of function. However, it must be noted that the conservation of active site residues suggests a conservation of mechanism, for example peptidase activity, and not necessarily conservation of substrate specificity. The involvement of peptidases with roles outside general amino acid metabolism as effectors in host–pathogen interaction is a more common phenomenon (Orbach et al., 2000; Xia, 2004).

The conservation of disulphide spacing in Vice5 and Vice10 with similar proteins in the public domain from S. sclerotiorum and P. nodorum may suggest an important role for these cysteine residues in protein function and/or stability within the apoplast (Joosten et al., 1994, 1997). Apart from Vice5 and Vice10, significant blastp similarity to P. nodorum proteins (i.e. within the top three blastp hits) was observed for the candidate effector protein sequences Vice3 and Vice6. This similarity reflects the recent release of P. nodorum protein sequences into the public domain, following completion of genome sequencing for this phytopathogen and the relatively close taxonomic relationship of P. nodorum to V. inaequalis (both Order Pleosporales; Goodwin, 2004). In the future, the use of similarities between candidate effectors and sequences present in the public domain will undoubtedly become more significant when more gene and protein sequence data from closely related organisms are released.

The predicted GPI modification sites of Vice6, Vice9 and Vice11 would, on initial consideration, preclude them from being successful effectors as, in the V. inaequalis–apple interaction, effectors are assumed to be soluble and secreted to the subcuticular space, or even taken up by the host cell, and these effectors are predicted to be covalently linked to the fungal plasma membrane or to β‐1,6‐glucans present within the fungal cell wall (de Groot et al., 2003; Richard and Plaine, 2007). However, some GPI‐anchored proteins are involved in host–pathogen interactions (Albrecht et al., 2006; Jiang et al., 2006; Richard and Plaine, 2007). For example, the majority of elicitins from Phytophthora, which are capable of inducing an HR in tobacco, possess a GPI anchor (Jiang et al., 2006). The question remains of how the plant detects these proteins to initiate a defence response when the corresponding proteins are not secreted to the plant–pathogen interface, although they may be released by cleavage by plant proteases.

The amino acid sequences of both Vice9 and Vice10 contain incomplete imperfect repeat domains (results not shown). Repeated amino acid domains have already been identified in two secreted proteins of V. inaequalis (Kucheryava et al., 2008). Secreted fungal proteins with internal repeats are predominantly involved in cell wall organization, and are often found coating the outer cell wall, participating in cell‐to‐cell or cell‐to‐substrate adherence or rendering the cell wall hydrophobic (Dranginis et al., 2007; Hung et al., 2002; Li and Palecek, 2008; Linder and Gustafsson, 2008; Richard and Dujon, 2006; Teerstra et al., 2006; Verstrepen et al., 2005; Wang and St. Leger, 2007). This is interesting given that Vice9 is predicted to possess a GPI modification site.

InterProScan results corroborated the blastp homologies and also revealed additional domains present within Vice11 and Vice14. The candidate effector Vice11 showed similarity to a C‐type lectin motif following the InterProScan searches. This motif is characteristic of C‐type lectins, which are not commonly found in either fungi or plants, although similar domains have been found in nine other fungal taxa to date, including two filamentous fungal plant pathogens Ustilago maydis and P. nodorum. However, there is a precedent for the involvement of a protein domain previously thought to be exclusive to the animal kingdom to be involved in a plant–pathogen interaction. The Avr4 chitin‐binding lectin effector of C. fulvum contains a chitin‐binding domain (CBM14) previously considered to be restricted to invertebrates. CBM14 protects the pathogen from the action of host chitinases, enhancing virulence (van den Burg et al., 2006). The candidate effector Vice14 showed similarities to domains associated with allergen V5/Tpx‐1, a PR protein (PR‐1) and SCP (sperm‐coating glycoprotein)‐like extracellular protein. These domains are all associated with extracellular eukaryotic proteins and are presumed to be evolutionarily related, but their precise function is unclear. Similarity to PR proteins is intriguing, given that these proteins are host encoded and contribute to resistance; however, a common structure does not necessarily indicate a common function. Proteins with domains similar to these are found in numerous fungal taxa, including hypothetical proteins in six filamentous fungal plant pathogens.

The SAM‐T06 server (http://www.cse.ucsc.edu/compbio/SAM_T06/T06‐query.html; Karplus et al., 1998, 2001, 2003, 2005) was employed to identify similarity between the predicted secondary and tertiary structure of the candidate effectors and solved structures. Results for Vice3 and Vice14 were consistent with blastp search results and functional motif database searches. For Vice10, weak structural similarities were identified to expansins, expansin‐like proteins, endoglucanases and pollen allergens. These motif signatures, together with the RlpA‐like double‐psi β‐barrel motif identified in blastp searches, are found within ATEXPA8 (GenBank ID: NP_181593), an α‐expansin of A. thaliana. Expansins induce plant cell wall extension (Choi et al., 2006). This result is interesting, given that a fungal endoglucanase has been identified with cell wall extension activity (Yuan et al., 2001).

Inconsistent with the observation that most cloned fungal effector genes encode cysteine‐rich proteins, the predicted protein sequence of effector candidate Vice6 possesses no cysteine residues. However, this candidate effector should not be discounted, as a paucity of cysteine residues has been demonstrated for PWL2 of Magnaporthe oryzae (Sweigard et al., 1995), AvrL567‐A and AvrM of M. lini (Catanzariti et al., 2006; Dodds et al., 2004) and AvrLm1 of Leptosphaeria maculans (Gout et al., 2006). In addition, five oomycete effector proteins are not cysteine rich (Allen et al., 2004; Armstrong et al., 2005; van Poppel et al., 2008; Rehmany et al., 2005; Shan et al., 2004).

The EST libraries from infected apple material and in vitro‐grown V. inaequalis have provided a valuable resource from which several candidate effectors have been mined. Of the 16 candidates analysed, the most interesting with respect to further investigation are Vice4, Vice12 and Vice16. These candidates fit the paradigm of filamentous fungal effectors in that they are small proteins with an even number of cysteine residues. Two are novel; the third (Vice16) is homologous to a hypothetical protein from a plant pathogen. All possess a signal peptide signature, without a transmembrane domain, and so are putatively secreted to the host–pathogen interface. Their expression at the transcriptional level is up‐regulated during growth in planta compared with that during growth in vitro. RNA‐mediated gene silencing (Fitzgerald et al., 2004) of these candidates will reveal any effector function and will contribute to the understanding of the V. inaequalis–apple interaction.

EXPERIMENTAL PROCEDURES

Fungal material and culture conditions

Venturia inaequalis isolates MNH120 (race 1, Auckland, New Zealand, Malus × domestica‘Granny Smith’, 1996) and MNH135 (race 1, Waikato, New Zealand, Malus × domestica‘Pink Lady’, 1996) were grown on cellophane (325P‐23 cellophane, Innovia Films, Melbourne, Vic., Australia; Parker et al., 1995) overlying potato dextrose agar (PDA; Oxoid Ltd., Basingstoke, Hampshire, UK) at 20 °C for 7–10 days (16 h light period/day) under white fluorescent lights (4300K, Cool white Deluxe, Sylvania, Danvers, MA, USA) for the production of conidia. Conidia for the inoculation of plants or media were harvested from agar plates by dislodging with SDW passed through a Milli‐Q Plus system (Millipore Corporation, Bedford, MA, USA), followed by filtration through sterile glass wool to remove mycelial debris; the concentration of conidia was adjusted to 1 × 10⁵ mL⁻¹.

Mycelia were cultured in liquid PDB (Difco Laboratories, Sparks, MD, USA) at 20 °C either with shaking at 150 r.p.m. (Orbital shaker, Sanyo Gallenkamp, Leicester, UK) following inoculation with 1 mL of conidial suspension, or statically as for PDA cultures. Mycelium (21–30 days old) was harvested by filtering through Miracloth (Calbiochem^®, La Jolla, CA, USA), and rinsed with 500 mL SDW prior to RNA or DNA extraction.

Cellophane sheets were removed from PDA plates at 7 dpi with V. inaequalis conidia, placed on the surface of 25 mL PDB and incubated as above for PDA cultures for 14–21 days. Cellophane sheets were rinsed in 20 mL SDW and blotted with 3M paper (Whatman Ltd., Maidstone, Kent, UK) to dislodge conidia prior to RNA extraction.

Plant material and infection assays

Seed originating from open pollinated Malus × domestica‘Royal Gala’ (Hawke's Bay, New Zealand) was used to produce infected plant material. Seeds were stratified for 6–8 weeks at 4 °C in fine‐grain vermiculite (Nuplex, Auckland, New Zealand) saturated with SDW, following surface sterilization for 1 min in Thiram (Kiwicare Corporation Ltd., Christchurch, New Zealand; 100 mg/mL fungicide). Following germination, seeds were grown for 4–6 weeks at 21 °C with a 12 h light period/day under 4C W SHP‐TS lights (Sylvania).

For library construction, seedlings were spray inoculated with a conidial suspension or SDW until run off using a painter's airbrush (Paasche Airbrush Co., Chicago, IL, USA). Seedlings were incubated in small Heat’n’Grow growth chambers (Hydroponic Wholesalers, Auckland, New Zealand) covered in aluminium foil. High humidity was maintained by adding 500 mL of 37 °C water and placing the chambers on heated plant‐raising panels (Hydroponic Wholesalers) for 24 h in a 20 °C room. After 72 h the foil was removed and the infection was left to develop at 20 °C under white fluorescent lights (4300K, Cool white Deluxe, Sylvania) with a 12 h light period/day. Leaves were harvested at 17 dpi for RNA preparation.

A detached leaf assay was used for qRT‐PCR (Win et al., 2003). Infected and SDW‐inoculated leaves were harvested at 5 and 10 dpi and used either for RNA extraction or microscopic evaluation of infection.

Microscopic evaluation of infection

Leaves were cleared and stained by the method of Bruzzese and Hasan (1983), and mounted in water on glass slides. Observations were made under bright field using a Research Photo‐micrographic microscope system (Olympus, Melville, NY, USA) with a 35‐mm automatic SLR camera attached. Digital images were captured using RS Image software (Roper Scientific Inc., Tucson, AZ, USA). Final figures were produced using Adobe^® Photoshop^® 6.0 (Adobe Systems Inc., San Jose, CA, USA.).

RNA extraction

Total RNA was extracted by the method of Chang et al. (1993) and concentration quantified using a Nanodrop^® ND‐1000 Spectrophotometer (NanoDrop Technologies, Rockland, DE, USA). mRNA was purified from total RNA using an mRNA purification kit (Amersham BioSciences UK Ltd., Little Chalfont, Buckinghamshire, UK), according to the manufacturer's instructions.

cDNA library construction

RNA was used as follows in the construction of the different libraries.

Infected leaf library

Total RNA was extracted from leaf material at 17 dpi with V. inaequalis MNH135 and used for library construction according to Newcomb et al. (2006). mRNA purified from this total RNA was ligated into λZAP^® (Stratagene, La Jolla, CA, USA), packaged into Gigapack III Gold (Stratagene) and plated using the host bacterial strain XLI‐BLUE MRF′ (Stratagene), according to the manufacturer's instructions. The library was mass excised from 250 000 plaque forming units (pfu) using ExAssist phage (Stratagene), according to the manufacturer's instructions. The transfection was performed on a large scale and plated onto 137‐mm Hybond™ N⁺ membranes (Amersham BioSciences UK Ltd.) overlaying Luria–Bertani (LB) agar. After 16 h, colonies were lifted from the membranes onto fresh Hybond N⁺. The original membrane and new replica membranes were then placed on LB agar plates containing 50 µg/mL ampicillin, and incubated at 37 °C for 5 h. After incubation, the original membrane was stored at 4 °C until needed and the replica membrane was cultured on LB containing ampicillin 170 µg/mL at 37 °C overnight. Blots were then denatured, neutralized and washed according to standard protocols (Sambrook et al., 1989) prior to being cross‐linked on a UV light box for 2 min. Blots were washed overnight in 50 mm tris(hydroxymethyl)aminomethane (Tris), pH 8.0, 1 m NaCl, 1 mm ethylenediaminetetraacetic acid (EDTA), 0.1% sodium dodecylsulphate (SDS) at 50 °C with shaking, and then wiped clean of bacterial debris prior to hybridization.

Genomic DNA was extracted from 21‐day‐old PDB shake cultures of V. inaequalis according to Win (2003), and sheared by a 10‐s burst of sonication (Kontes, Vineland, NJ, USA) prior to use as a hybridization probe to select for fungal sequences according to standard protocols (Sambrook et al., 1989). Colonies corresponding to hybridizing spots were picked and then mixed with 500 µL SM buffer (0.1 m NaCl, 10 mm MgSO₄, 50 mm Tris‐HCl, pH 7.5) and left overnight at 4 °C. PCR using T3 and T7 universal primers was carried out with 2 µL of the plaque/buffer as template. Cycling parameters were as follows: 94 °C for 2 min, followed by 35 cycles of 94 °C for 10 s, 58 °C for 30 s, 72 °C for 30 s, and then 72 °C for 10 min. PCR‐amplified inserts were purified using PCR purification columns (Qiagen GmbH, Hilden, Germany) prior to sequencing.

In vitro‐grown V. inaequalis libraries

Total RNA extracted from 30‐day‐old mycelium from PDB shake cultures was used for library construction. mRNA packaged into plaques and the preparation of sequencing template were performed as detailed above for the infected leaf library.

SSH library enriched for cellophane‐induced genes

mRNA from 28‐day‐old statically grown mycelium and cellophane‐grown V. inaequalis MNH120 was used in the construction of an SSH library employing the PCR‐select™ cDNA subtraction kit (BD Biosciences Clontech, Palo Alto, CA, USA), according to the manufacturer's instructions, with cellophane‐grown V. inaequalis cDNA as tester and the statically grown mycelium cDNA as driver.

An indication of subtraction efficiency in the MAAD library was given by comparing the prevalence of cell function genes in the in vitro libraries quantified following a keyword search using ‘actin’, ‘ribosomal protein’, ‘glyceraldehyde 3‐phosphate dehydrogenase’ (gapdh), ‘ubiquitin’ and ‘ubiquitin conjugating enzymes’ versus the annotations assigned to each EST based on a blastp algorithm (Altschul et al., 1997) search of the NCBI RefSeq (http://blast.ncbi.nlm.nih.gov/Blast.cgi; Pruitt et al., 2007) and UniProt UniRef90 (http://www.ebi.ac.uk/uniprot/; Suzek et al., 2007) databases.

DNA sequencing and bioinformatics

Sequencing of the clones from the in planta libraries was carried out at Genesis Research and Development, Auckland, New Zealand, and clones from the in vitro libraries (including the SSH library) were sequenced at the School of Biological Sciences Sequencing Unit, University of Auckland, New Zealand. Big Dye Terminator standard protocols were used (Applied Biosystems, Foster City, CA, USA) and sequencing reactions were resolved on ABI377, ABI3100 or ABI3700 sequencers according to the manufacturer's instructions (Applied Biosystems). The ESTs were either automatically or manually trimmed of vector, adapter and low‐quality sequence regions, and entered into a relational database. ESTs were assembled into non‐redundant sequences using cap3 (using a 95% level of identity as a threshold; Huang and Madan, 1999), and automatic annotation was performed using the HortResearch BioPipe sequence annotation pipeline (a cluster‐based annotation system written in PERL; R. Crowhurst, M. Davy and C. Deng, The New Zealand Institute for Plant and Food Research Limited, Auckland, New Zealand, unpublished work) and utilizing a relational database (MySQL; http://www.mysql.com). Redundancy within the libraries was calculated using the formula: total number of ESTs—(total number of contigs + total number of singletons)/total number of sequences (Cramer et al., 2006).

The most likely full‐length ORF of each candidate effector was identified by distinguishing the longest ORF with a 5′ ATG with a high degree of conservation of the Kozak sequence specific to filamentous fungi, denoted by the consensus nucleotide sequence CA[CA][AC]AUGGC (Balance, 1986; Kozak, 1986). blastp algorithms (Altschul et al., 1997) were used to distinguish plant and fungal sequences from the in planta libraries. Those sequences that had an E‐value of ≤ 1e‐10 for their best match against fungal sequences present in the NCBI RefSeq (Pruitt et al., 2007) or UniProt UniRef90 (Suzek et al., 2007) databases were assigned a fungal origin. Similarly, for sequences showing a higher similarity (< 1e‐10) with plant sequences, a plant origin was assigned. For those sequences with less significant E‐values (≥ 1e‐10 but ≤ 1e‐5), assignments of origin were still made as above. Any sequences that showed no significant identity with known sequences were also assigned a fungal origin to avoid missing putative novel fungal sequences. To ensure that each of the candidate effector proteins was predicted to be secreted, the most probable full‐length amino acid sequence encoded by each candidate effector gene was screened for the presence of a putative N‐terminal signal peptide using the SignalP 3.0 server (http://www.cbs.dtu.dk/services/SignalP/; Bendtsen et al., 2004; Nielsen et al., 1997). A eukaryotic‐specific server setting was employed and run according to the neural networks (NN) and hidden Markov models (HMM). To determine whether any of the predicted proteins were targeted to subcellular localizations other than the secretory pathway, the TargetP 1.1 server (http://www.cbs.dtu.dk/services/TargetP; Emanuelsson et al., 2000, 2007) was used with default cut‐off and non‐plant organism settings. The presence of transmembrane domains was determined by analysis using the TMHMM server v. 2.0 (http://www.cbs.dtu.dk/services/TMHMM/; Krogh et al., 2001; Sonnhammer et al., 1998).

In an attempt to estimate which cysteine residues are predicted to be involved in disulphide bond formation, the disulphide bonding state and connectivity pattern for each cysteine residue within each candidate effector protein were analysed by the DISULFIND server (http://disulfind.dsi.unifi.it/help.php; Ceroni et al., 2006).

Candidate effector protein sequences were screened for the presence of putative C‐terminal GPI lipid modification (anchor) sites using the big‐PI server, developed for the assessment of fungal protein sequences known to enter the endoplasmic reticulum–Golgi secretory pathway as predicted by signal peptide prediction software (http://mendel.imp.ac.at/gpi/fungi_server.html; Eisenhaber et al., 2004).

Common protein domains and functional motifs within these proteins were identified using the InterProScan database server (http://www.ebi.ac.uk/Tools/InterProScan/; Mulder et al., 2007; Quevillon et al., 2005).

Based on a combination of protein fold recognition, sequence alignment and several methods of local secondary structure prediction, the SAM‐T06 server was employed to identify similarities between candidate effector‐predicted secondary and tertiary structure and proteins with solved structures (http://www.cse.ucsc.edu/compbio/SAM_T06/T06‐query.html; Karplus et al., 1998, 2001, 2003, 2005). An expected value threshold of > 1e‐05 was enforced for similarities identified on the basis of structure (unless similarities were to proteins already known to play a role in infection specificity in other fungi).

Genomic DNA extraction and PCR analysis

The extraction of genomic DNA from isolate MNH120 was carried out according to Win (2003), following 4 weeks of growth in PDB. Genomic DNA was quantified using a Nanodrop^® ND‐1000 Spectrophotometer (NanoDrop Technologies). For candidate effector genes of ≤ 1 kb in length, genomic DNA amplifications were performed within a final reaction volume of 20 µL containing 2 ng/µL genomic DNA template, 300 nm of each primer (Table 5), 1 U Platinum Taq DNA polymerase (Invitrogen, Carlsbad, CA, USA), 200 µm deoxynucleoside triphosphates (dNTPs) (Invitrogen), 1.5 mm MgCl₂ and 1 × PCR buffer (Invitrogen). For genes > 1 kb in length, genomic DNA amplifications were performed within a final reaction volume of 20 µL containing 2 ng/µL genomic DNA template, 300 nm of each primer (Table 5), 1 U Expand Long Template Taq + Tgo DNA polymerase mix with proofreading activity (Roche Diagnostics GmbH, Mannheim, Germany), 200 µm dNTPs (Invitrogen), 1.5 mm MgCl₂ and 1 × PCR buffer (Invitrogen). The cycling parameters were as follows for templates ≤ 1 kb : 95 °C for 5 min, 35 cycles of 94 °C for 10 s, 60 °C for 1 min, 72 °C for 1 min, followed by 10 min at 70 °C, and a 15 °C hold. This was modified for templates > 1 kb in length to have a 3‐min extension time at 72 °C. PCR products were resolved by agarose gel electrophoresis.

Table 5.

Primers used in this study.

Gene name	Primer name	Primer sequence (5′−3′)	Usage
Vi‐ gapdh	Vi‐GAPDH‐F	GGCAAGACCATCCGTTTCTA	Assessing purity of cDNA
	Vi‐GAPDH‐R	GACACCCATGACGAACATTG	Assessing purity of cDNA
Md‐ gapdh	Md‐GAPDH‐F	TGGAAAATTGACCGGAATGT	Assessing purity of cDNA
	Md‐GAPDH‐R	GACCTGCTGTCACCAATGAA	Assessing purity of cDNA
Vice1	Vice1AF	TCCTCACCACAACCCTCAC	qRT‐PCR
	Vice1R	CGAAGTTACCACCAGCAACA	qRT‐PCR
	Vice1gF	GGCTGGAACTGCTGGCTC	Genomic DNA PCR
	Vice1gR	CTATACGCCCTGTTTGCA	Genomic DNA PCR
Vice3	Vice3F	ACCCAATCAACGGACCTCTC	qRT‐PCR
	Vice3R	CCCTTCAACCATTCGCCTAC	qRT‐PCR
	Vice3gF	CAAGCTCCCTTGCAGCAAC	Genomic DNA PCR
	Vice3gR	TCAAAACTGTGCCAGGATG	Genomic DNA PCR
Vice4	Vice4F	TGAAGCTCCTACTTATTCCCATC	qRT‐PCR
	Vice4R	GTATCGACATCGCCTGCTTT	qRT‐PCR
	Vice4gF	GGCATGTAAAT	Genomic DNA PCR
	Vice4gR	CTATCCCTTCCCACAT	Genomic DNA PCR
Vice5	Vice5RTF	TCGGAAAGGACGGTGACTAC	qRT‐PCR
	Vice5RTR	AGACGGTTTTGGTGCTCTTG	qRT‐PCR
Vice6	Vice6F	TCGCAGTCACCTCCCAAC	qRT‐PCR
	Vice6R	TTTGAAGCCGAGATAATGGA	qRT‐PCR
	Vice6gF	GCTCAAGACGCATCCTCAGCA	Genomic DNA PCR
	Vice6gR	CTACAAGAAAGCAGCAAAC	Genomic DNA PCR
Vice7	Vice7AF	TGGTGATGGCTCCTACACAG	qRT‐PCR
	Vice7AR	CACATCCGTTTCCGTTTACC	qRT‐PCR
Vice8	Vice8RTF	ATTTGCCAGGCATACAAAGC	qRT‐PCR
	Vice8RTR	AGGAGGACGAGAGTGATTGC	qRT‐PCR
	Vice8gF	ATGATCCTCGAGCCCAAAG	Genomic DNA PCR
	Vice8gR	TCATAGGTCTTCTTTGAATAG	Genomic DNA PCR
Vice9	Vice9RTF	ACAGCTGTGGATGGGAAGAA	qRT‐PCR
	Vice9RTR	TGGGTGCCAACCACCTAC	qRT‐PCR
	Vice9gF	GCTGAGTATTACGATTA	Genomic DNA PCR
	Vice9gR	TTAGAAGAGAGCAGCAAC	Genomic DNA PCR
Vice10	Vice10F	AGGAATGGGTGACGGAGTG	qRT‐PCR
	Vice10R	CCCTGGTCTTGGTGCTTG	qRT‐PCR
Vice11	Vice11F	TGCCATCATTGGTATTCAAG	qRT‐PCR
	Vice11R	AAGATTTGGGTCAGCACAGG	qRT‐PCR
Vice12	Vice12AF	GCACATTCCCAACCCTCGAG	qRT‐PCR
	Vice12R	CAGGAACCGACAAGAAGAGC	qRT‐PCR
Vice14	Vice14F	CAAGGACTGGAAGCACAAGA	qRT‐PCR
	Vice14R	CGTTGTAGGCTGCGTTTGTG	qRT‐PCR
Vice16	Vice16CF	GTCCTGTGGCTTGGTCTTTG	qRT‐PCR
	Vice16CR	AATGGTTGCCCGAGTGGT	qRT‐PCR
β‐ Tubulin	Vi‐TUB‐F	CGTCGTGAGGCTGAAGGT	qRT‐PCR
	Vi‐TUB‐R	CGATGGGACAACAGAGAATG	qRT‐PCR
ribo L12	Vi‐L12‐F	GTTGTCCCATCTGCCTCTTC	qRT‐PCR
	Vi‐L12‐R	GTCCTTGGCCATTGACTTGT	qRT‐PCR

Md‐gapdh, Malus glyceraldehyde 3‐phosphate dehydrogenase; qRT‐PCR, quantitative reverse transcriptase‐polymerase chain reaction; ribo L12, 60S ribosomal protein L12 (EU853840); β‐tubulin, EU853839; Vice, Venturia inaequalis candidate effector; Vi‐gapdh, Venturia inaequalis glyceraldehyde 3‐phosphate dehydrogenase.

qRT‐PCR

Infected and SDW‐inoculated leaves were harvested at 5 and 10 dpi, and 10‐day‐old PDB shake culture in vitro‐grown mycelia of V. inaequalis were used for RNA preparation as above. RNA was treated with DNAse I (Invitrogen) according to the manufacturer's instructions to remove any contaminating genomic DNA prior to cDNA synthesis. Absence of gDNA was confirmed by PCR using primers specific for V. inaequalis gapdh (EU873167) and apple gapdh (Table 5). cDNA was synthesized from this RNA using the Transcriptor first strand cDNA synthesis kit (Roche Diagnostics GmbH) according to the manufacturer's instructions; cDNA from two separate syntheses was pooled to control for cDNA synthesis bias. One‐in‐ten dilutions of cDNA in SDW were used for qRT‐PCR.

qRT‐PCRs were carried out on a Lightcycler^® 480 instrument (Roche Diagnostics GmbH) using a SYBR Green detection system. The default cycling programme was used with the following modifications: a 62 °C annealing temperature for 10 s and a 6 s extension time. The primers used for qRT‐PCR are shown in Table 5. Data were gathered from two independent infection/growth assays, each treatment within the assays being represented by three biological samples (each sample comprising three infected leaves or material gathered from three flasks of PDB), and the PCR for each biological sample was conducted in triplicate. PCR efficiency for each set of primers was calculated using Lightcycler^® 480 internal software. A gene expression normalization factor was calculated for each sample based on the geometric mean of two stably expressed reference genes (β‐tubulin; EU853839) and 60S ribosomal protein L12 (EU853840) using geNorm software v3.4 (Vandesompele et al., 2002), and used for the calculation of the relative expression of each targeted gene in each sample.

REGISTRATION OF SEQUENCES

Sequences relating to the candidate effectors can be found in the GenBank database. Vice1, EB148895; Vice2, EB149687; Vice3, EB149718; Vice4, EB149885; Vice5, FJ621507; Vice6, EB152934; Vice7, EB153294; Vice8, EB153275; Vice9, EB153319; Vice10, FJ621508; Vice11, FJ621509; Vice12, FJ621510; Vice13, FJ621511; Vice14, FJ621512; Vice15, FJ621513; Vice16, FJ621514.