A proprietary resource of c. 50,000 perennial ryegrass expressed sequence tags (ESTs) 46 was integrated into the Bioinformatic Advanced Scientific Computing (BASC) system 47. Each EST was functionally annotated using data from microarray-based transcriptomics experiments, the rice Ensembl browser, Pfam and gene ontology databases. BASC was used to search for the presence of NBS-LRR sequences. A text search with the query terms 'disease' and 'resistance' was used to identify candidates based on a wuBLASTX threshold of E = 10^-15 through known gene ontology within the genomes of closely-related cereal species (wheat, oat and barley), rice and Arabidopsis.

Locus amplification primers (LAPs) for multiple target genes were designed using standard parameters as previously described 48. LAPs were designed from perennial ryegrass EST templates, and sequence tagged site (STS) primers derived from Italian ryegrass (L. multiflorum Lam.) NBS sequences located in GenBank 39.

LAPs were designed based on the sequence of four oat LGB-located Pca cluster R genes 37, five barley rust resistance genes (Hvs-18, Hvs-133-2, Hvs-T65, Hvs-236 and Hvs-L6) 33; and the third exon and 3'-terminus of the TaLrk10 extracellular domain 41.

Degenerate primers (4 in sense and 12 in antisense orientation) were designed to the conserved regions (P-loop and GLPL) of cloned oat R genes 37 and were used in conjunction with published R gene-specific degenerate primers 33343849 (Additional File 1). Based on interpretation of initial amplicon complexity, specific primers were subsequently designed for SNP discovery.

For specific homologous and heterologous R gene-derived primers, PCR amplicons were generated using template genomic DNA from the parental genotypes of the F₁(NA₆ × AU₆) mapping population 4850. For degenerate primers, genomic DNA from the crown rust resistant Vedette₆ genotype 14 was used as an primary template, and re-designed primer pairs were used with the F₁(NA₆ × AU₆) parents. Amplicons were cloned and sequenced essentially as previously described 48, except that a total of 32 Vedette₆ clones and 12 clones from each of NA₆ and AU₆ were analysed. Trace sequence files were used as input materials into the BASC module ESTdB 47.

All candidate NBS-LRR (R gene) nucleotide sequences were subjected to two-way BLASTX and wuBLASTX analysis against the GenBank and the Uniprot databases, respectively. Genomic DNA sequences were translated to amino acid sequences using Transeq software. Each peptide sequence was scanned against the Pfam database 5152 for the presence of known domains, the type, size and position of NBS domains and the number of LRR repeats. Multiple Expectation Maximisation for Motif Elicitation (MEME) 53 was used to detect conserved motifs between sequences containing NBS domains 34.

Preliminary alignments of predicted protein sequences was performed manually using Bioedit (version 7.0.5.3 – Ibis Biosciences, Carlsbad, CA, USA). The alignments were split into two separate datasets (for the P-Loop to GLPL region, and for the Kin-2A to GLPL region), and were realigned for phylogenetic analysis using CLUSTALX 54 with default options. Clustering of related sequences based on amino acid homology was conducted using a Neighbour Joining (NJ) algorithm and bootstrap analysis was calculated on an unrooted NJ cladogram based on 1000 iterations using CLUSTALX 55.

Perennial ryegrass genomic DNA was extracted from parents and progeny of the F₁(NA₆ × AU₆), Vedette₆ and p150/112 4556 mapping families using the CTAB method 57. A genotypic panel for genetic map assignment was constructed of 141 F₁(NA₆ × AU₆) and 24 p150/112 F₁ genotypes as previously described 21.

PCR-amplified genomic amplicons were cloned and sequenced and DNA sequences were aligned essentially as previously described 48. Predicted SNPs were initially validated using 10 F₁(NA₆ × AU₆) individuals, and those showing Mendelian segregation were then genotyped across the full mapping panel through the single nucleotide primer extension (SNuPe) assay 48. Integration of SNP loci into the existing F₁(NA₆ × AU₆) parental genetic maps was performed as previously described 214850.

Comparison of chromosomal regions controlling crown rust resistance between perennial ryegrass trait-specific mapping populations was performed using data from QTL analysis of the F₁(SB2 × TC1) mapping population 17. The F₁(SB2 × TC1) parental maps contained heterologous RFLP and genomic DNA-derived SSR (LPSSR) markers shared with the p150/112 and F₁(NA₆ × AU₆) genetic maps, respectively 4556. Comparison of marker locus order between the p150/112 and F₁(NA₆ × AU₆) genetic maps was performed through the presence of common LPSSR loci 5056. This common marker set also allowed interpolation of the position of the LpPc1 crown rust resistance locus on p150/112 LG2 14. Chromosomal locations of LrK10 ortholoci were compared between Lolium and Avena species using common heterologous RFLP loci 58. Further comparative genomic analysis was conducted using published genetic maps from cereal species including barley, wheat, rye and oat 3359.

Three strategies (empirical approaches based on heterologous PCR and degenerate PCR, and a bioinformatic discovery method) resulted in the identification of 67 primary R gene templates for host genetic analysis (Table 1). Initial PCR amplification and resequencing using the parental genotypes of the F₁(NA₆ × AU₆) mapping population allowed identification of a further 35 secondary R gene template sequences (Additional File 2). A total of 14 primer pairs amplified paralogous sequences, at a mean of 2.5 per primary template sequence, with a range from 1–12. A total of 102 distinct putative R gene sequences (corresponding to 99 NBS-containing genes and 3 receptor kinase genes) were annotated (Additional File 2) and subjected to further characterisation. Representative genomic sequence haplotypes were deposited as accessions for unrestricted access in GenBank (accession numbers FI856066–FI856167). A schematic summary of the candidate gene discovery process and further applications is depicted in Figure 1.

In the empirical approach category, translational genomics between perennial ryegrass and closely related cereal species (oat, barley and wheat) which are susceptible to other Puccinia rust pathogens (P. coronata f. sp. avenae, P. hordii, P. triticina) was used to identify R genes. Perennial ryegrass amplicons derived from oat R gene template primer pairs demonstrated high BLASTX similarity matches to their corresponding template sequences (data not shown). Primer pairs designed to the TaLrk10 template generated two 1.6 kb fragments, one of which (LpLrk10.1) displayed very high similarity scores to the putative oat ortholocus (AsPc68LrkA).

The specificity of amplification using degenerate primers designed to amplify NBS domains was dependent on the proportion of deoxyinosine (I)-containing nucleotides. Those based on oat R gene templates contained a high frequency of inosines (>15%) and predominantly amplified retrotransposon-like sequences (data not shown). In contrast, combinations of largely non-degenerate primer pairs based on sequence information from multiple Poaceae species (barley, sorghum and ryegrass) (Additional File 1), successfully generated NBS domain-containing amplicons of the correct size (Additional File 3). A total of 28 distinct NBS domain-containing sequences (Tables 1, Additional File 2) were generated, several primer pairs generating multiple products (up to 7) (Additional File 3).

The text search-based computational approach identified 23 distinct perennial ryegrass ESTs with high sequence similarity to known resistance genes from closely-related species (Table 1, Additional File 2). Amplification based on candidate EST primary templates was efficient, with only 13% of LAP pairs failing to generate amplicons. Additional sequences were amplified from several ESTs, all were putative paralogues showing significant BLASTX similarity (E < 1 × 10^-15) to known R genes (Additional File 2).

Database searches for previously-characterised ryegrass NBS sequences identified 51 accessions from Italian ryegrass-derived clones and a further 14 from an interspecific L. perenne × L. multiflorum hybrid (L. x. boucheanum). All 6 previously-described STS primer pairs successfully generated single amplicons of the expected size (Table 1, Additional File 2).

From the total of 102 analysed sequences, 89 (87%) exhibited BLASTX matches at E < 10^-20 to known NBS domain-containing sequences from closely-related cereal species in both the GenBank and UniProt databases (Additional File 2). In most cases (80%), the highest matching sequence was the same for both databases. Sequence translation and subsequent Pfam analysis revealed that a substantial proportion of partial protein sequences were similar to the NBS domain (Additional File 4). A large proportion of the NBS-category sequences (55%) were within the NBS domain, while the remaining sequences either overlapped the NBS region at the N- or C- terminus, contained the LRR domain, or were located solely within the N- or C- terminal domain. A range of different R gene sub-classes containing NBS, CC-NBS, NBS-LRR, NBS-NBS-LRR, CC-NBS-LRR, CC-CC-NBS-LRR and NBS-NBS domains were detected, but no TIR-NBS containing sequences were observed. Of the different sub-classes of NBS sequences, 52 contained 1–33 LRRs (modal at 3), 25 contained 1 or more CC domains, and five sequences contained the NBS-NBS domain (Additional File 4). A further three receptor kinase and NBS-LRR genes contained trans-membrane domains.

Consensuses were determined for the seven major NBS domain motifs (P-Loop, RNBS-I, Kin-2A, RNBS-II, RNBS-III, GLPL and RNBS-V) (Additional File 5) and were compared to those from closely related Poaceae species (wheat and rice) and to A. thaliana. The P-Loop, Kin-2A and GLPL motifs were most highly conserved between all species examined, while the RNBS-I and RNBS-II motifs were conserved within the Poaceae, and the Kin-2A and RNBS-II motifs were the most conserved among the CC-NBS sequences. The RNBS-III and RNBS-V motifs were highly divergent between all species.

A total of 50 different motif signatures were identified by MEME analysis with 60 NBS domain-containing sequences at an average of 13 residues in length. The most commonly occurring signatures were components of the conserved regions such as the P-Loop, Kin-2A and GLPL motifs (Fig. 2, Additional File 6). All the distinct sub-classes of NBS sequences present either completely lacked, or contained highly variable RNBS regions. Structural analysis revealed substantial diversity in motif content within the NBS domain and grouping of specific motifs into sub-classes based on shared sequence origin.

Phylogenetic analysis was performed based on two selected NBS domain regions (P-Loop-GLPL and Kin-2A-GLPL). Unrelated NBS domain sequences from A. thaliana, lettuce (Lactuca sativa L.), flax (Linum usitatissimum L.), tomato (Lycopersicon esculentum L.) oat, rice and barley were included for both regions, as were GenBank-derived Lolium NBS sequences. A total of 38 P-Loop-GLPL sequences and 104 Kin-2A-GLPL sequences were analysed. Amino acid alignment of NBS regions permitted classification into sub-families or classes. A total of seven major clusters were identified for the P-Loop-GLPL region (Additional File 7, Additional File 8). Candidate sequences were clustered on the basis of similarity to putative orthologues identified from preliminary BLASTX analysis. The majority were most closely related to those from other ryegrass species, although some showed highest sequence similarity to template genes from other species. Sequences similar to rice R genes were also grouped with flax, lettuce and A. thaliana accessions [cluster C], and a sub-set of ryegrass sequences formed two separate clusters [clusters G and H] and may hence be similar to generic R gene variants previously identified in other species, which were not included within the alignment.

Eight major clusters were identified for the Kin-2A-GLPL region (Additional File 9, Additional File 10). Ryegrass-derived sequences were preferentially clustered with those from other Poaceae species (for instance, with oat sequences formerly used as LAP-design templates [cluster A], and with rice and barley sequences [clusters C and G, respectively]). Sequences from a number of dicotyledonous plant species were separately clustered for the P-Loop-GLPL [cluster E], but co-located in several distinct clusters [cluster D and E] with ryegrass-derived sequences for the Kin-2A-GLPL region.

Sixty-five distinct R gene templates were subjected to in vitro SNP discovery through resequencing from parental genotypes of the F₁(NA₆ × AU₆) mapping population. Genomic DNA of a cumulative length of c. 37 kb was analysed and a total of 819 R gene SNPs were predicted, at an overall frequency of 1 per 46 bp. A total of 11 (17%) template biparental contigs contained no SNPs, while 27 (42%) of the remaining templates contained under 10 SNPs (Table 2). All monomorphic R gene contigs were derived from the NBS domain, apart from two encoding receptor kinase-like enzymes. SNP incidence was low within introns, due to limited representation in the sample set. SNP frequencies within parental genotypes was higher for NA₆ (38) than for AU₆ (20). A further 8 SNPs with biparental (AB × AB) segregation structures and 4 SNPs with AA × BB structures were identified.

Multiple R gene SNPs from 37 (69%) of 54 SNP-containing R gene contigs were validated (Additional File 11). A total of 26 R genes were assigned to loci on the parental maps of the F₁ (NA₆ × AU₆) mapping population (22 on all NA₆ LGs [Figs. 3, 4], 10 on all but LG4 for AU₆ [Figs. 5, 6]). SNPs in four R gene loci showed biparental segregation structures, mapping to the equivalent LG position in each parental map, and hence provide bridging markers. Five loci were also mapped to equivalent positions on three p150/112 LGs. A single SNP locus derived from the template sequence LpHvESTClone1.1 (xlprg50-464ca) was mapped in p150/112 but not in F₁(NA₆ × AU₆) (Fig. 7)

R gene locus clusters were identified on a number of LGs, often in close proximity to mapped DR gene loci (represented by SNP and previously mapped EST-RFLP loci). Major clusters were identified in the lower regions of LGs 1 and 2 and the upper region of LG5 of both F₁ (NA₆ × AU₆) parental maps (Fig. 3, 4, 5, 6).

Genetic mapping facilitated map integration between trait-specific ryegrass genetic maps, and also comparative relationships with other Lolium and Poaceae taxa. Coincidences between SNP loci assigned to the F₁(NA₆ × AU₆) parental maps and crown rust resistance QTLs detected in other studies were observed for LGs 1, 2, 5, and 7. Two R gene loci co-located with the crown rust resistance QTLs LpPc2 and LpPc4 in the lower region of LG1 (Fig. 8). A further two loci were assigned to the centromeric region of p150/112 LG2, 4 cM distant from the genomic DNA-derived SSR locus xlpssrk02e02 which is closely associated with LpPc1. This marker locus group also co-locates with LpPc3 in the F₁(SB2 × TC1) LG2 map, and through comparative alignment, with the hexaploid oat Pca cluster on LGB based on the position of the heterologous RFLP locus xcdo385.2 (Fig. 9). The R gene SNP locus xlprg60-216gt mapped adjacent to a previously-identified crown rust resistance QTL on AU₆ LG7, and in putative alignment with a corresponding QTL on LG7 of the Lolium interspecific hybrid ψ-F₂(MFA × MFB) population map, but a limited number of common markers precluded further interpretation (data not shown).

Comparative genomic analysis detected conserved relationships between perennial ryegrass Lrk10 R gene SNP locus (xlprg1-369ct) and the corresponding cereal LrK10 template genes. A macrosyntenic region was identified on LG1, although low numbers of common genetic markers again limited the accuracy of extrapolation (Fig. 10). The perennial ryegrass R gene loci xlprg24-460at and xlprg54-688ag are derived from putative orthologues of the barley R genes HvS-217 and HvS-L8, respectively. Alignment of genetic maps revealed conserved syntenic locations, as well as coincidence with QTLs for leaf rust and powdery mildew resistance on barley 2H and 3H, respectively (Additional File 12, Additional File 13).

This study describes the most comprehensive study to date of ryegrass NBS domain-containing sequences. The largest comparable surveys were of R genes from Italian ryegrass (62 sequences: 39) and from both annual and perennial ryegrass and the corresponding interspecific hybrid (16 sequences: 38, all derived by means of degenerate primer-based amplification. In this study, 102 distinct R genes were isolated and functionally annotated. Bioinformatic analysis identified the majority of candidate genes as members of the NBS-LRR family responsible for major gene resistance in plant species 296061626364. A proportion of c. 20% of all perennial ryegrass R genes may be estimated to have been sampled, assuming equivalent gene content to that revealed (545 NBS sequences) by the genome-wide survey of rice 31. It is possible, however, that major rounds of genome duplication or divergence events between species may have occurred, based on different selection pressures of surrounding pathogen populations. Such factors may influence the relative number of NBS-containing sequences in ryegrass species.

Results from the current study suggest that only non-TIR NBS sequences are present within the Lolium genome, consistent with previous results from monocotyledonous species 333738395865. Only degenerate primers specific to non-TIR sequences were able to amplify PCR products from perennial ryegrass genomic DNA, as observed in similar studies of sorghum 34.

Substantial variation was observed within coding regions of non-TIR NBS-LRRs, which exhibit greater sequence diversity than the TIR-NBS sub-family 66. In this study, many R genes lacked the P-Loop region, while others contained NBS-NBS domains, duplicated CC regions or lacked CC and/or LRR domains. P-Loop, Kin-2A and GLPL motifs were conserved and similar in sequence to those of closely related Poaceae species such as wheat and rice 3158 and the model dicotyledonous species A. thaliana 30. Further evidence for structural gene diversity was observed within particular NBS sub-families. NBS sub-classes contained specific signature motifs between conserved regions, and in some instances, RNBS motifs were missing or duplicated. This suggests that the RNBS-I and RNBS-II motifs may either play a role in pathogen-specific recognition, or be less functionally significant than other, more highly conserved domains mediating resistance in plant species 3066. Alternatively, the presence of CC-NBS-specific motifs may suggest divergence to perform specialised functions. Variability was also observed within LRR domains, suggesting that NBS-LRRs in ryegrass are diverse in function 6467.

Amino acid diversity in the P-Loop-Kin-2A region may account for the major differences between TIR-NBS and CC-NBS domains. The results from this study demonstrate that TIR-NBS sequences from flax and A. thaliana group in a separate cluster, as observed in a previous phylogenetic analysis of Lolium NBS domains 38. Further sequence analysis of a larger number of Lolium sequences in the Kin-2A-GLPL motif interval demonstrated increased sequence similarity with known TIR-NBS regions from dicotyledonous plant species, suggesting that this region may be more conserved across taxa. Consensus motif order and sequence composition indicates that the Lolium RNBS-I region may have diverged from that of dicotyledonous plants. Similar results were observed in other Poaceae species such as sorghum, for which RNBS-I consensus sequences showed significantly higher similarity to those of rice than to those of A. thaliana TIR-NBS genes 34.

Phylogenetic analysis of the P-Loop-GLPL and the Kin-2A-GLPL domains detected at least 8 NBS sub-classes, as compared to 5 separate clusters identified in a previous study 38. Analysis of the larger number of Kin-2A-GLPL interval sequences obtained only one more cluster than for the P-Loop-GLPL interval, indicative of domain conservation. Inclusion of NBS sequences from other closely, and more distantly related, species permitted grouping of R genes and inference of possible common origins for R gene sub-families. Sequences amplified from oat templates clustered together with ryegrass template-derived R genes, suggestive of a common origin. Based on known mechanisms of R gene evolution, gene duplication and divergence prior to speciation within the Pooideae sub-tribe is likely to account for the sequence similarity between ryegrass and oat genes, corresponding to putative orthologues 496168.

The SNP frequency observed within this study was marginally lower than that detected within a sub-set of 11 perennial ryegrass R genes across 20 diverse genotypes 69, but similar to that observed within DR genes 21 and a broad range of functionally-annotated candidate genes 48 in the F₁(NA₆ × AU₆) mapping population. Eight R gene templates contained up to 90 SNPs per contig, possibly due to paralogous sequence alignment. Large numbers of haplotypes have been reported for other perennial ryegrass NBS-LRR genes, especially within variable LRR regions 69. The data from this study suggests that allelic diversity within NBS domain is low compared to the highly variable LRR domain.

Previous studies identified significantly non-random chromosomal distributions of NBS-containing sequences 3031: 44 gene clusters were detected in the japonica sub-species of rice. Five major clusters containing two or more closely linked NBS-LRR genes, which frequently showed low mutual sequence similarity, were identified from only a small sub-set (26) of mapped perennial ryegrass R genes. This suggests that the gene location pattern in perennial ryegrass may be similar to that observed in other plant species. Unrelated R genes also mapped in close association with DR gene SNP and RFLP loci 2150. QTL based analysis and genetic mapping in wheat identified co-location of DR and R genes at qualitative disease resistance loci 7071. Co-location of R genes with DR genes was also observed in similar chromosomal regions (lower regions of LG1, LG2 and LG6) as disease resistance QTLs which were mapped both in F₁(NA₆ × AU₆) and other trait-specific mapping populations 172021.

SNP mapping of two candidate R genes in both the F₁(NA₆ × AU₆) and p150/112 mapping populations has provided possible candidates for the major gene crown rust resistance QTL (LpPc1) on LG2 14. To determine whether R gene SNP variants are of functional significance, further experiments involving transgenic approaches, association genetic analysis or map-based cloning are required 7273.

NBS-LRR genes loci mapping to the distal region of LG1 in the F₁(NA₆x AU₆) parental genetic maps (xlprg29-293at, xlprg30-707ag and xlprg61-23ga) are potential candidates for resistance effects to crown rust pathotypes which are yet to be identified within Australasia. Major QTLs for crown rust resistance (LpPc2 and LpPc4) have been mapped to the lower part of LG1 in 3 different perennial ryegrass trait-specific mapping populations 171819 but the limited number of common markers limits accurate extrapolation between genetic maps. Two QTLs of large magnitude were identified in each of the three populations, which were screened using European crown rust isolates However, so far no resistance QTLs have been detected within this chromosomal region using isolates from the southern hemisphere. As both F₁(NA₆ × AU₆) mapping population parental genotypes are derived from Eurasia 50, LG1-located R gene polymorphisms may confer resistance to crown rust isolates of European provenance.

Comparative analysis of R gene SNP loci and corresponding ortholoci confirmed previously reported macrosyntenic relationships between perennial ryegrass and other Poaceae species 45 in nearly all instances. The sole exception was the xlprg8-271ct locus, which was assigned to LG1 despite being derived from (and highly similar to) an oat Pca template gene predicted to map to LG2. Genetic mapping of the LpLrk10 locus to LG1 suggested that the structure and chromosomal location of this gene are highly conserved throughout the Pooideae 414358. The equivalent analysis for barley R gene ortholoci provides the basis for testing R gene functionality in response to a broader range of plant diseases, requiring significant improvements of pathogen phenotyping 74 and corresponding genetic analysis 4275.