To identify genes related to early events of urediniospore germination and differentiation, we incubated freshly harvested urediniospores of P. striiformis f. sp tritici (race CY32) in sterile distilled water in plastic plates. After incubating for 10 h at 9°C, about 60% of urediniospores attached to the plastic surface and produced long, un-branched germ tubes (Fig. 1). Some of these germ tubes displayed various degree of swelling at the tip (Fig. 1). Germinated and un-germinated urediniospores were collected at 10 h and used for RNA isolation. A directional cDNA library consisting of 6.05 × 10⁵ primary clones was constructed with the λTripEx2 vector (Clontech). On average, about 95% of the cDNA clones had inserts longer than 0.2 kb. The insert size varied from 200–3500 bp, with an average of 750 bp. A total of 5500 random cDNA clones were sequenced from the 5'-end to obtain 4810 quality reads or ESTs.

Before clustering analysis, ESTs containing no insert sequence or insert shorter than 100 bp were removed. The remaining 4798 ESTs were aligned and assembled into 315 contigs (containing two or more ESTs) and 803 singletons. A total of 1118 unisequences (uniseqs) were submitted to dbEST at NCBI 18 under GenBank accession numbers ES321929 – ES323046. Most of the uniseqs had the insert size of 200–900 bp. Only 36 uniseqs were longer than 1000 bp. The G+C content of these ESTs ranged from 40.74% to 53.73%, with an average of 45.08%, which was similar to the G+C content of ESTs in P. graminis 12.

The number of individual ESTs belonging to each contig ranged from 2 to 608 [see Additional file 1]. Nine most abundant contigs contained more than 100 ESTs each, suggesting a high transcript level of the corresponding genes. Contig Ps303 was represented by 608 clones, more than any other contigs [see Additional file 2]. It is homologous to a Saccharomyces cerevisiae gene encoding a hypothetical protein. Five of these most abundant contigs [see Additional file 2] had no homologs in EST or protein database bases. Three other abundant contigs displayed limited homology with entries in GenBank. Contig Ps253 had weak homology with a putative secreted protein in Ixodes scapularis. Contig Ps28 consisting of 128 ESTs was homologous to a differentiation-related protein Inf24 from Uromyces appendiculatus. Inf24 was identified as infection structure protein that was highly expressed during urediniospore germination 19. For contig Ps314, it consisted of 154 ESTs and was homologous to a predicted protein of Kluyveromyces lactis.

All uniseqs were subjected to similarity searches against sequences in the non-redundant protein (nr) database at GenBank using the BLASTX algorithm. InterProScan was used to analyze uniseqs with no BLASTX matches for known protein domains. The vast majority (703) of the uniseqs had no significant homolog in GenBank by BLASTX search. Among 415 uniseqs displayed similarity (E-value < 10e-5) to entries in the nr database, about 56% of them were similar to genes coding for proteins with unknown functions (not classified in Gene Ontology). The frequency of orphan sequences in P. striiformis f. sp. tritici ESTs is similar to that has been observed in other fungal EST projects 20.

Among the 415 uniseqs that exhibited similarity to entries in the non-redundant protein database at GenBank, about 40.72% shared homology to proteins from filamentous fungi while the rest were homologous to proteins from a wide variety of organisms, including yeast, bacteria, nematodes, plants, insects, and animals [see Additional file 3]. The uniseqs with significant similarity to hypothetical proteins were placed in the unclassified protein category. Based on the results from BlastX and InterProScan searches, ESTs with significant matches were categorized according to their putative functions (Table 1, for detail [see Additional file 4]).

A total of 267 uniseqs were identified to be homologous to proteins with known cellular functions. Figure 2 listed the putative functions of these uniseqs and their occurrence. Over 70% of them were functionally related to primary metabolism (42.59%) and protein (21.58%) or (11.11%) RNA synthesis. About 10% were involved in cellular signaling and defense responses. The ESTs with no hits in GenBank were searched against the genome sequence of the wheat stem rust P. graminis 21. Many of them had homologous sequences in the genome of the stem rust fungus (data not shown). A total of 195 uniseqs had no homologous sequences in the stem rust fungus genome. These ESTs may represent genes unique to P. striiformis f. sp. tritici.

To date, many genes important for fungal pathogenesis have been identified in plant and human pathogens 222324. Homology searches revealed that several ESTs were homologous to known fungal pathogenicity or virulence factors (E-value < 1e-5), including key components of signal transduction pathways, transporters, and genes involved in infection-related morphogenesis (Table 2). For three genes related to signal transduction, Ps1728 encodes a putative adenylate cyclase that has been shown to be essential for pathogenicity in Magnaporthe grisea, Ustilago maydis, and several other fungi 25262728. Ps259 encodes a putative Gα subunit that is similar to MagB of M. grisea and Gpa1 of Cryptococcus neoformans 2930. The MAP kinase encoded by Ps261 is homologous to Pmk1 of M. grisea, which is important for regulating plant infection processes in a number of phytopathogenic fungi 313233.

Ps8 and Ps28 may be related to the infection structure differentiation (Table 2). Both of them were homologous to Inf24 of U. appendiculatus and appeared to be members of a multigene family. Inf24 is highly expressed during urediniospore germination. Microinjection of antisense oligonucleotides of Inf24 inhibits appressorium development in U. appendiculatus 19. Both Ps28 and Ps228 were contigs represented by over 100 ESTs, indicating that these two genes are highly expressed in germinated urediniospores, and may have similar functions during early stages of plant infection in P. striiformis f. sp. tritici.

Contigs Ps6 and Ps159 encoded proteins homologous to gEgh16 of Blumeria graminis 22, which is highly expressed at early infection stages (16 h). In M. grisea, its homologs, GAS1 and GAS2, are important virulence factors that are involved in appressorial penetration 34. Both GAS1 and GAS2 are specifically expressed during appressorium formation. We also identified two contigs, Ps300 and Ps5709, that were putative ATP-binding cassette (ABC) transporter genes. In M. grisea and other plant pathogens, several ABC transporter genes have been implicated in fungal-plant interactions, possibly by controlling the efflux of phytotoxic fungal metabolites or plant defensive compounds 35.

We also identified an EST, Ps238 (Table 2), that encodes a putative copper- and zinc-superoxide dismutase. In some fungal pathogens, such as Candida albicans, superoxide dismutase is required for tolerance to oxidative stress and full virulence 36. Contig Ps85 encoded a protein consisting of 169 amino acid residues. It is homologous a putative cell surface antigen gene in C. albicans 37 and contains a CFEM (

C

F

E

M

The first Avr protein identified in rust fungi was AvrL567 of the flax rust fungus Melampsora lini 40. Recently, a number of additional avirulence elicitors have been identified in the flax rust fungus 41. In the ESTs of the wheat stripe rust fungus, we identified a candidate avirulence elicitor encoded by contig Ps87, which had significant homology with HESP767. In the flax rust fungus, HESP767 is expressed in haustoria and may function as an effector in interactions with the host plant 41. Ps87 contained the full length open reading frame (ORF) that was predicted to encode a protein of 85 amino acid residues. It had a typical signal peptide at the N-terminus (data not shown) as predicted by SignalP 3.0 42. Similar to HESP767, Ps87 may function as an effector or avirulence factor during wheat infection in P. striiformis f. sp. tritici.

Six uniseqs, Ps28 (a differentiation-related protein), Ps85 (a CFEM domain-containing protein), Ps159 (gEgh16 homolog), Ps259 (G-alpha subunit), Ps261 (Pmk1 homolog), and Ps87 (HESP767 homolog), were selected for assaying their transcript levels during different developmental and infection stages by quantitative real-time PCR (qRT-PCR). All these six genes were expressed in urediniospores, germinated urediniospores, and infected wheat tissues harvested at 1 through 7 days post inoculation (dpi). However, their transcription levels were much higher in germinated urediniospores than in un-germinated urediniospores or infected wheat tissues (Fig. 3; Fig. 4).

Although the transcript level of these genes in infected wheat tissues was relatively low, they had different expression patterns during plant infection. In comparison with the others, Ps159 had a relatively high transcript level in infected wheat plants. Its expression was up-regulated from 2 to 6 dpi and then slightly decreased at 7 dpi. The transcript level of Ps85 and Ps28 increased at early infection stages but decreased after 5 dpi. The transcript level of Ps259 and Ps261 in wheat tissues was lower than that of the other ESTs examined. Those two genes may be specifically and highly expressed during urediniospore germination. These data suggest that a considerable reprogramming of gene expression occurs during urediniospore germination, early stages of plant infection, and the biotrophic growth in P. striiformis f. sp. tritici.

During incompatible interactions, Ps87 also displayed an expression pattern that decreased at early stages but slightly increased after 3 dpi (Fig. 4). The high transcript level of Ps87 in germinated urediniospores suggests that it may be involved in germ tube growth and appressorium differentiation. It is possible that Ps87 also plays a role in late stages of plant infection, probably for host-pathogen recognitions.

P. striiformis f. sp. tritici strain CY32 was inoculated and propagated on wheat cultivar Huixianhong as described previously 57. Fresh urediniospores were harvested from infected wheat plants and resuspended in sterile distilled water (6 mg/200 ml). After incubating in a 50 × 125 cm dish at 9°C for 10 h, germinated urediniospores were collected with a spatula, frozen in liquid nitrogen, and stored at -80°C. For isolating RNA from infected plants, wheat leaves of susceptible cultivar Huixianhong and resistant cultivar Shuiyuan 11 were inoculated with CY32 urediniospores and harvested at 1, 2, 3, 4, 5, 6, and 7 days post inoculation (dpi).

Total RNA was isolated from 100 mg of germinated urediniospores with the RNeasy Plant Mini RNA purification kit (QIAGEN, Germany) following the instruction provided by the manufacturer. The SMART^TM cDNA library construction kit (Clontech, USA) was used for cDNA synthesis and library construction. The MaxPlax^TM Lambda Packaging Extract (Epicentre, USA) was used for in vitro packaging and transfection of Escherichia coli strain XL1-Blue. The resulting directional library consisting of 6.05 × 10⁵ primary clones was amplified and stored in 15% glycerol at -80°C. Randomly selected cDNA clones were sequenced with primer seq1 (5' CGACTCTAGACTCGAGCAAG 3') from the 5'-end with an ABI 3130-XL DNA sequencer.

Sequence reads longer than 100 base pairs were processed by cross_match 58 that allows the removal of contaminated sequences. Repeats and low complexity sequences were masked using RepeatMasker 59. The resulting quality trimmed sequences were extracted and assembled with the CAP3 assembler, and viewed with Consed 60. Statistics of the assemblies were generated by perl scripts using BioPerl modules.

For functional classification, the resulting unisequences were searched against the NCBI non-redundant protein database using the BLASTX program 6162. The unisequences with significant BLASTX matches were classified according to their likely cellular functions following the general categories outlined by the Gene Ontology Consortium 63. InterProScan was used to search for protein domains in ESTs with no significant homologs in BLASTX searches.

Standard protocols 64 were used to extract total RNA from urediniospores, germinated urediniospores, and wheat leaves inoculated with the stripe rust fungus. Two micrograms of total RNA each was used for cDNA synthesis with the SuperScript First-strand Synthesis System (Invitrogen) following the instruction provided by the manufacture. The resulting 1^st-strand cDNA products synthesized with RNA from urediniospores, germinated urediniospores, and infected wheat leaves were used as the templates for qRT-PCR assays.

Four sets of primers and fluorescently labeled TaqMan probes [see Additional file 5] were used to assay the transcript levels of ESTs Ps28, Ps85, Ps159, Ps259, Ps261, and Ps87. The expected sizes of resulting PCR products were 107, 65, 81, 73, 103, and 179 bp for Ps28, Ps85, Ps87, Ps159, Ps259, and Ps261, respectively. The fluorogenic probes were labeled at the 5'end with the fluorescent reporter dye FAM (6-carboxy-fluorescein) and modified at the 3'end with the quencher dye TAMRA (6-carboxy-tetramethylrhodamine). Transcript abundance was assessed with three independent biological replicates. Each real-time PCR mixture (25 μl) contained 12.5 μl premix (Premix Ex TaqTM, Takara), 7.5 μl distilled H₂O, 1 μl 10 μM forward primer, 1 μl 10 μM reverse primer, 1 μl TaqMan probe, and 2 μl the cDNA template. The Smart CyclerSystem (Cepheid, USA) was used for PCR reactions that consisted of 10 s at 95°C (1 cycle) and 40 cycles of 95°C for 5 s and 60°C for 20 s. The transcript level of selected ESTs was calculated by the 2^-ΔΔCT method 65 with the actin gene of P. striiformis f. sp. tritici as the endogenous reference for normalization. Relative quantification of these ESTs was computed with their transcript levels in different stages in comparison to that in urediniospores.