Plants have evolved a sophisticated, multi-layered defence network to detect and respond to pathogen challenges. Inducible responses are governed by plasma membrane pattern recognition receptors (PRRs) and also cytoplasmic immune receptors encoded by resistance (R) genes. PRRs recognise relatively conserved small molecules, proteins and protein fragments, produced externally to the cell by invading pathogens, and collectively referred to as pathogen associated molecular patterns (PAMPS). By contrast, R proteins directly or indirectly perceive proteins and small molecules termed effectors that are introduced into plant cells by the pathogen. Genes encoding effectors that are recognised by R gene products, leading to effective plant resistance, are genetically defined as avirulence (avr) genes. Two modes of resistance may be distinguished: PAMP triggered immunity (PTI) that is mediated by PRRs, and effector triggered immunity (ETI) that results from effector recognition by R proteins and often produces a hypersensitive response, a form of localised host programmed cell death [1]. R genes have been implicated in resistances against diverse and taxonomically unrelated pathogens including bacteria, viruses, nematodes, insects, filamentous fungi and oomycetes. In addition to being pivotal for host resistance, PRRs and R genes are thought to play a role in non-host resistance [2].

The majority of cloned and functional R genes described within the plant kingdom contain a nucleotide-binding site (NB) and leucine-rich repeat (LRR) domain, and are members of the STAND (Signal Transduction ATPase with Numerous Domains) protein family of NTPases, known as NB-LRRs [3, 4]. The nucleotide binding site forms part of a larger complex known as NB-ARC, which reflects its presence in the human apoptotic protease-activating factor-1 (APAF-1), plant R proteins and Caenorhabditis elegans death-4 protein (CED-4) [5]. Further subdomains and multiple conserved motifs have been identified within the NB-ARC domain [3].

Based on the presence or absence of N-terminal domains, members of the NB-LRR family can be divided into two major groups. The first group contains an N-terminal domain with homology to the Drosophila toll and human interleukin-1 receptor (TIR) and is referred to as TIR-NB-LRRs or TNLs. The second, non-TIR-NB-LRR, group is collectively known as CNLs as some, but not all, members of this group contain a predicted coiled-coil (CC) structure in the N-terminus. This division of NB-LRR proteins is also reflected in phylogenetic analyses of the NB-ARC domains in which TNL and CNL proteins form distinct clades [6–8].

NB-LRR genes comprise one of the largest gene families in plants. Approximately 150 NB-LRR encoding genes have been identified in the genome of Arabidopsis thaliana Col-0 [9], 185 within Arabidopsis lyrata[10], 92 within Brassica rapa[11], 416 and 535 in the genomes of the woody species poplar and grapevine respectively [12], and 464 and 483 in two genomes of Oryza sativa[13]. In addition, partial NB-LRRs that lack some NB-LRR specific domains and contain, for example, only TIR, TIR-NB, CC, and CC-NB domains, have been described in plant genomes [8, 10]. NB-LRR genes are ancient in their origin and have been identified in ancestors of early land plants. NB genes with sequence homology to TNLs have been described in bryophytes [14] and TNLs and CNLs have been found in gymnosperms and eudicots [15]. However, the composition of NB-LRR genes varies significantly between species [16]. The unequal representation of NB-LRR lineages within plant taxa has been typified by the low frequency of TNLs within the monocotyledonous species despite the manifestation of TNLs prior to the angiosperm-gymnosperm split [15, 17].

Within genomes, NB-LRR genes are organized either as isolated genes, or as linked clusters of varying size that are thought to facilitate rapid R gene evolution [18]. NB-LRR gene clusters are termed homogeneous when they contain only sequences that share a recent common ancestor. In contrast, clusters that contain more distantly-related NB-LRRs are referred to as heterogeneous [19].

Potato is the most important non-cereal food crop, with worldwide production yielding approximately 330 million tonnes in 2009 (http://faostat.fao.org/site/339/default.aspx). Like all plants, potato faces a constant barrage of pest and microbial threats. More than 50 functional NB-LRR genes have been cloned from potato and related members of the Solanaceae [20] and 738 NB-LRR-like sequences have previously been identified in a BAC library prepared from a heterozygous diploid potato clone, RH [21]. The genome sequence of a doubled monoploid Solanum tuberosum group Phureja clone, DM1-3 516 R44 (hereafter referred to as DM), has recently been described [22]. Among the 39,031 annotated protein coding genes, 408 NB-LRR coding genes were predicted. In this study we used a process of iterated computational and manual annotation to further identify potential NB-LRR coding sequences, determine their locations on the 12 potato chromosomes and study the phylogenetic and positional relationships between the individual genes. Our results provide significant insight into the evolution of NB-LRRs and, importantly, a blueprint for future efforts to identify and more rapidly clone functional NB-LRR genes from Solanum species.

We used an iterative process of manual and computational analysis to identify 438 NB-LRR-encoding sequences within the recently published doubled monoploid potato genome [22]. This study has revealed a slightly higher number of CNLs and TNLs compared to the 408 NB-LRRs described by Xu et al. (2011) [22]. The difference, which is within the expected sampling error, includes 28 additional TNL genes and 2 additional CNLs. By extending the DM gene models by 3 kb at the 3' and 5' end respectively to produce the DMG+ sequences, more domains associated with NB-LRR type genes were identified and the gene annotations correspondingly extended. The number of annotated partial NB-ARC only genes fell in our predictions from 105 to 41 (Table 2). Whilst our analysis used NB-LRR discriminative MEME motifs derived from a training set harbouring functionally characterised NB-LRRs from the wider plant kingdom, the analysis described by Xu et al. (2011) [22] is based on NB-derived Pfam domain searches, followed by the construction of a potato-specific NB hidden Markov model. Both approaches yielded very similar numbers of NB-LRRs. Unfortunately, a direct comparison between the different resistance gene homologues (RGHs) was not possible as the identities of the CNL and TNL genes predicted by Xu et al. (2011) [22] were not made publicly available.

The MEME motif and phylogenetic analysis revealed a distinction between CNLs and TNLs in the N-terminal region, and in the NB-ARC domain of these sequences. Seven of the 20 identified MEME motifs (Table 1) distinguished between these NB-LRR subclasses, or between the canonical and RPW8-type CNLs [24]. Phylogenetic analysis, which was performed on the conserved NB-ARC domain, supported this distinction and was consistent with previous observations for other plant species [6, 7, 9–12, 24, 49].

The DM potato genome harbours 4.7 times more CNL than TNL genes. A similar distribution was found for the NB-LRR genes of grapevine (3.8×), but the ratio is smaller in poplar (1.7×) [12]. In comparison, the NB-LRRs of the Brassicaceae A. thaliana, A. lyrata and B. rapa contain CNLs and TNLs in a 1:2 ratio [9–11]. The genome of the monocot rice contains only CNLs; all other grasses analysed so far contain no or only very few TNLs [13, 15]. Leister (2006) suggested that overrepresentation of TNL over CNL genes in the Brassicaceae Arabidopsis and rape seed could reflect the adaptation of the R gene set to the predominant pathogens. It can be speculated that the over-representation of CNLs in potato is a response to some of the most damaging pathogens such as P. infestans, which is typically controlled by CNLs. In line with this, it is interesting to note that 27% of the identified NB-LRR genes share high sequence similarity to functionally characterised Rpi genes (data not shown).

The proportion of all genes that are predicted to encode NB-LRRs is 1.16%, which is in line with estimates for other plant species that range between 0.6-1.8% [11]. The gene density around potato NB-LRR loci is approximately 100 genes per megabase. However, unlike RxLR effectors from P. infestans which often reside in gene sparse regions [50], a global analysis of the DM NB-LRR genes (Figure 5) shows that CNLs and TNLs reside in genomic regions that are not significantly different to the potato genome in general in terms of gene or repeat density.

The CNL branch forms two phylogenetic clades, containing the canonical CNLs and the CNL-R (CC_RPW8-type), as previously described [24, 49]. Within the canonical CNLs, eight major subgroups with high support and short branch length were identified, suggesting a recent common ancestor. Two-thirds (13 of the 21) of the functional CNL genes included in the tree are found in these subgroups. Only members of CNL-3 and CNL-7 (and some of the smaller subgroups) show no significant sequence similarity to a functionally characterised R gene thus far. Their role, which is hitherto unknown, could for example be to provide resistance to yet unknown pathogens and/or to mediate non-host resistance responses [2].

Several approaches for the identification of NB-LRR clusters have been described elsewhere, and we have utilised a combination of the analyses described by [12] and [9]. The identified members and the overall number of predicted clusters were very similar for both types of analyses, suggesting that the identification of clusters by these methods is relatively robust. However, cluster prediction based on the distances between NB-LRRs does not take into account the variability of gene density in the potato genome [22]. Similarly, the definition of a gene cluster solely based on the number of non-NB-LRR genes between CNLs and TNLs fails to take any physical distance into account. Predicted potato NB-LRR genes are unevenly distributed over the 12 chromosomes and cluster into groups of different sizes. This is in line with data for other plant species [9, 11, 12]. Various mechanisms including recombination, gene conversion, duplication and selection are thought to contribute to the genome-wide diversity and distribution of NB-LRR gene loci [19, 51–53]. Equal intragenic crossing-over results in domain swaps between genes whereas unequal crossing-over influences the number of genes within a locus and potentially places genes into a new structural context. Tandem duplications, in which the copy is contiguous to the original gene, are typically associated with homogeneous clusters. Of the 63 clusters, 50 are homogeneous and thus likely a result of tandem duplications. Members of the subgroups CNL-1 to CNL-8 are often found on the same chromosome and, in some cases, within the same clusters, which is consistent with tandem duplication. In contrast, segmental and ectopic duplications, which involve the duplication of entire gene blocks or single/small groups of genes respectively, can position copies to unlinked sites including different chromosomes [51]. Both CNL and TNL distributions display evidence for events that placed homologous genes onto different chromosomes that could be a result of either segmental or ectopic duplication. These events appear to be more common for TNLs that are more widely dispersed throughout the genome and not found in clusters as frequently as CNLs.

The sequencing of DM provides a snapshot of the potato genome organisation, and specifically the distribution of and relationships among NB-LRR genes on individual chromosomes. Although specific to DM, this analysis provides important insight into the NB-LRR gene compositions of other members of the Solanaceae. Studies in Arabidopsis have shown, for example, that some R genes display high levels of polymorphism within and between populations [10]. A more detailed analysis of the potato R1 locus [46], for which three haplotypes from S. demissum have been described [47], confirmed evidence of copy number variations and is consistent with tandem duplications. As previously described, the R1 locus is flanked by Bs4-like and Prf-like genes but the number of R1-homologues varies between one and 17 in S. demissum and five in DM (Figure 4c; [47]). Another example is the R3 locus on chromosome 11 which was originally described in a diploid potato population, SHxRH [54]. Overall, R3 cluster organisation is syntenic between SH-haplotypes and the sequenced DM, in that the R3a-clusters (C52, C53, and C54 proximal) and the R3b cluster (C55, distal) flank the marker GP185 (Figure 4b). However, in DM, the physical distance between the clusters C54 and C55 amounts to more than 350 kb and is thus approximately 200 kb shorter than the same region in SH [43]. In DM, nine R3b homologues reside in cluster C55, and Li et al. (2011) [43] describe six and ten homologues for the two SH haplotypes. Unequal representation of lineages within the NB-LRR superfamily and copy number variation between haplotypes is consistent with a 'birth and death' model in which some NB-LRRs are lost and new lineages evolve whilst others are retained [55].

We have observed 438 NB-LRR genes in a doubled monoploid potato, which represents a single haplotype. Potato cultivars and breeding lines are generally heterozygous tetraploids, which exhibit tetrasomic inheritance during crossing. The high levels of structural diversity observed in homologous R gene clusters from different potato haplotypes (e.g. [21, 46, 47]), and the extremely high levels of sequence polymorphism observed in potato, imply that it is highly likely that any given tetraploid potato clone may contain as many as 1,600 distinct NB-LRRs in its genome. A key objective for future resistance breeding is to understand the allelic diversity of NB-LRR genes in potato. Such an objective will require application of high throughput sequencing technologies allied to advanced bioinformatic tools for assembling sequence data from very closely related genes.

We have identified 438 NB-LRR type genes within the sequenced potato S. tuberosum Group Phureja (DM), of which several are homologous to functionally characterised R genes. Comprehensive analysis of the NB-LRRs, both in terms of the phylogenetic relationships of CNLs and TNLs and their positions on the respective chromosomes, provides an invaluable tool for the identification of novel and functional R genes from wild Solanum species in the future. New technologies, including exon capture followed by high throughput sequencing and allele mining rely on detailed information concerning R gene organisation and distribution. Furthermore, knowledge about the genomic organisation of these genes will facilitate comparative and evolutionary studies on a whole genome level or, alternatively, for selected clusters.