Plant resistance genes (R genes) play an important role in direct or indirect recognition of proteins encoded by specific avirulence (Avr) genes of pathogens 1 . In recent years, a large number of R genes which confer resistance to diverse pathogens have been cloned from different plant species by using either map-based cloning or transposon tagging approaches 2 3 4 5 . Based on the amino acid sequence comparison among cloned R genes, several highly conserved domains, such as NBS, LRR, protein kinase (PK), transmembrane domain (TM), and TIR were identified as the majority of known R genes 6 7 8 9 . These distinct conserved domains provide scientists a convenient way to identify and then clone additional R genes and RGLs or resistance gene analogs (RGAs).

The largest class consists of R genes which contain NBS and LRR domains (NBS-LRR). This class can be further divided into two sub-classes, TIR-NBS-LRR and non-TIR-NBS-LRR, based on the presence or absence of a TIR domain at the N-terminus 7 10 . The second class of R genes is comprised of those with LRR and PK domains (LRR-PK), such as Fls2 in Arabidopsis and Xa21 in rice 11 12 . The third class contains those with a large extracellular LRR domain, for example, Cf-4 in tomato and Vf1 in apple 13 14 . The fourth class contains those only with the PK domain, such as Pto in tomato and Pbs1 in Arabidopsis 15 16 . And finally, the fifth class includes all other R genes that confer resistance through different mechanisms, for example, Hm1 in maize and mlo in barley 3 17 .

Based on conserved domains of known R genes, PCR amplification has been successfully used to isolate RGLs or RGAs from Arabidopsis 18 19 , soybean 20 21 22 , rice 23 , maize 24 25 , wheat 26 , and many other plant species 27 28 29 30 . For example, Bertioli et al. isolated 78 RGAs from cultivated and wild peanuts by using degenerate primers designed on NBS conserved domains 31 . Leal-Bertioli et al. identified candidate genome regions for disease resistances by using a combination of NBS profiling, sequence characterized amplified region (SCAR) markers, and Southern blot by probes with NBS encoding regions 32 . In common bean, eight classes of disease resistance related sequences and 15 RGAs have been isolated based on the conserved NBS and TIR domains 33 34 . An RGA, SB1, was found to cosegregate with the Crg gene which is known to be responsible for resistance to the rust pathogen Uromyces appendiculatus 35 .

Recently, data-mining approaches have been employed to identify RGLs and RGAs from plant EST databases. In maize, 199 RGLs have been identified using data-mining, modified amplified fragment length polymorphism (AFLP), and rapid amplification of cDNA ends (RACE) methods 25 . In sugarcane, Rossi et al. identified 88 R-gene-like ESTs via the data-mining method 36 . In wheat, Dilbirligi and Gill found 220 expressed R-gene candidates using modified RNA fingerprinting and data-mining approaches 37 . In Medicago, Zhu et al. have revealed 179 unique NBS-LRR RGLs by using combination methods of database queries, hybridization with RGLs from related species, and PCR 38 . With the availability of several complete and nearly complete plant genome sequences, R gene analyses at the whole genome level have been carried out in Arabidopsis 10 , rice 39 , Medicago 40 , and Lotus 41 . As a result, 149 NBS-LRR-encoding genes and an additional 58 related genes have been identified in the Arabidopsis genome 10 . Also 489 predicted NBS-LRR genes were identified in the rice genome, all of which belong to the non-TIR-NBS-LRR class 39 . From the Medicago truncatula draft genome (Mt1.0), Ameline-Torregrosa et al. have identified 333 non-redundant NBS-LRRs and predicted 400 to 500 NBS-LRRs in the full genome 40 . In the Lotus genome, 158 NBS-encoding sequences have been isolated and nearly 40% of them are pseudogenes encoding incomplete protein sequences 41 . With advances in sequencing technologies over the past few years, next-generation sequencing technologies have become widely available and have dramatically accelerated biological research 42 , making identification of R gene homologs and members of other related gene families by data-mining methods much more efficient.

Common bean (Phaseolus vulgaris L.) is one of the most important legume crops in the world. Due to its high protein and low fat content, it has become a principal source of dietary protein and fiber in many developing countries, especially those in Eastern and Southern Africa and Latin America 43 . Diseases such as anthracnose, angular leaf spot (ALS), rust, common bacterial blight (CBB), and Bean golden yellow mosaic virus (BGYMV) highly reduce bean production and quality. For example, CBB is a serious seed-borne disease that can cause over 40% yield loss in both temperate and tropical bean production regions 44 . Rust can cause 18% to 100% yield loss of susceptible bean genotypes under favorable environmental conditions 45 . Therefore, it is very important to better understand disease resistance in common bean.

The objective of this study was to identify expressed RGLs from common bean and to develop RGL-tagged molecular markers for genetic mapping of candidate genes and QTLs for disease resistance.

As a component of the gene-for-gene resistance mechanism in plants, resistance genes play an important role in recognizing products encoded by specific avirulence genes of a pathogen 49 . In this study, 365 tentative RGLs from common bean were successfully identified by data-mining based on the availability of 454-derived sequences in our lab and common bean ESTs in GenBank. About 60% (218) of identified PvRGLs were from 454-derived sequences; moreover, 166 (76.15%) of the 218 PvRGLs were new transcripts.

ESTs are highly valuable for genome annotation and gene structure prediction 50 . The 454 sequencing is a faster and more cost-effective method of producing sequence data than the Sanger method and is capable of producing a 400 to 600 million base pairs per run with 400 to 500 base pair read length 51 . It has been successfully used to maximize gene discovery, improve gene predictions, and detect SNPs and mutations 52 53 . In the past few years, data-mining approaches have been successfully used to isolate RGLs or RGAs from sugarcane 36 , wheat 37 and maize 25 . As reported previously, Dilbirligi and Gill 37 adopted four different data-mining methods including domain search, individual and multiple motif searches, consensus sequence search, and individual full-length search to mine R-gene-like wheat sequences, and showed that the individual full-length search was the most successful method. There were 243 NBS-LRRs in addition to 101 other types of expressed R-gene candidates which were then isolated via an individual full-length search using a low E value of e-1 37 . Xiao et al. 25 used three methods including modified AFLP, RACE and data-mining to isolate RGAs and R-gene-like ESTs from maize and found that the data-mining method is the most efficient way. A total of 186 expressed RGAs were recovered from 550,000 maize ESTs using a moderate E value of e-10 or better 25 . Rossi et al. revealed 88 RGLs from sugarcane ESTs by using a very stringent E value of e-50 or better and represented three major classes of R genes, namely NBS-LRR, LRR-TM and PK 36 . The above three research reports showed that different E values have a great effect on the number of resulting RGA or RGLs. In the present study, the moderate E value of e-10, similar to that in maize, was used to mine common bean RGLs and a total of 365 tentative PvRGLs were identified. Of the 365 tentative PvRGLs, 29 belonged to NBS-LRR type, 96 belonged to LRR, LRR-TM, and LRR-PK type, 229 belonged to PK type, six and five contained sequences with similar to putative TM domains and Toxin reductase domains, respectively. The number of RGLs identified in the present study was about two times more than those in maize due to three reasons. Firstly, 1.77 million 454-derived sequences and common bean ESTs were screened to identify RGLs in common bean, about three times more EST sequences than those in maize (550,000 ESTs). Secondly, 454 sequencing can generate more sequence data than the Sanger sequencing. Therefore, transcripts at extremely low levels can be detected 51 . Finally, some identified PvRGLs match the same R genes or RGLs. For example, PvRGL083 and PvRGL236 match different regions of the same BAC-end sequence.

RT-PCR was used to examine the expression of the PvRGLs in this study. Results indicated that all of the selected PvRGLs were actually expressed in the leaves of genotype Sierra (Figure 1). In contrast, many RGLs or RGAs amplified from genomic DNA using degenerate primers or mined from whole genome are not expressed 33 . Previously, eight classes of disease-resistance related sequences were amplified from common bean DNA using degenerate primers based on the conserved NBS domain 33 . Expression analysis indicated that three RGAs (SB1, SB3 and SB8) were not expressed 33 . In Lotus, 62 NBS-encoding sequences were considered as pseudogenes due to encoding of incomplete protein sequences 41 . In Arabidopsis, at least 12 NBS-LRR-encoding genes were predicted to be pseudogenes due to frame shift and nonsense mutations 10 .

The PvRGLs discovered in this study correspond to most of the 25 previous common bean R genes or RGAs in the PRGdb database 2 . PvRGL266 and PvRGL275 have strong hits to the P. vulgaris TL5601 disease resistance protein gene with amino acid identity of 90% and 85%, respectively. TL5601, located in the I locus of common bean, controls resistance to Bean Common Mosaic Virus 54 . PvRGL262 matched to coiled-coil NBS-LRR (CNL)-B11 with amino acid identity of 94%. CNL-B11 was mapped in the B4 R gene cluster which contained at least three R genes (Co-9, Co-y, and Co-z) and QTL effective against anthracnose, and Bean golden yellow mosaic virus 55 56 . PvRGL309 was a part of polygalacturonase-inhibiting protein (PGIP) gene. PGIP can inhibit fungal endopolygalacturonases and is considered to be an important factor for plant resistance to phytopathogenic fungi 57 . PvRGL294 was the same as, but much longer in the 5' end than the previous RGL SB3 and OB9 33 .

So far, most of the known R genes have been cloned by map-based cloning and transposon tagging approaches 58 . Therefore, mapping RGLs and RGAs to genomes and/or genetic maps is very important and will facilitate R gene cloning. In this study, 105 PvRGLs could be integrated into the common bean FPC physical map by comparison of PvRGLs to P. vulgaris BAC-end sequences. Additionally, we were able to anchor 237 PvRGLs to the common bean genetic map by using conserved syntenic blocks between common bean and soybean. The PvRGLs are broadly, but unevenly distributed among the 11 linkage groups of common bean with a strong tendency of clustering. For example, 41 PvRGLs, the largest number, were mapped to Pv8, while 17 of them clustered at the bottom of Pv8. David et al. (2008) also found that many specific R genes against various pathogens cluster together: for example, in the B4 resistance gene cluster, 73 BAC clones (FI159954 - FI160067) were identified by using a NBS probe PRLJ1 to screen a common bean BAC library 56 . In the present study, PvRGL173 was anchored at the top of Pv4 by in silico mapping. Meanwhile, PvRGL173 shows high similarity (92.56% and 96.79%) with BAC-end sequence FI160023 (5e-153) and FI159996 (1e-101), respectively. PvRGL173 should be located in the B4 resistance gene cluster. The B4 resistance gene cluster contains various R genes resistant to different pathogens, such as Colletotrichum lindemuthianum (anthracnose), Uromyces appendiculatus (rust), and Pseudomonas syringae pv. phaseolicola (halo blight), in addition to QTLs for resistance to BGYMV and anthracnose; therefore further work will be needed to determine the role of PvRGL173. In Medicago, two superclusters of disease resistance genes were identified. One is located at the top of Mt3 containing 73 CNL and 9 TNL encoding sequences. The other is located at the bottom of Mt6 containing 57 TNL encoding sequences 38 . Clustering of R genes facilitates the genetic variation of R genes and benefits in the evolution of new R genes. Bertioli et al. found that retrotransposons were associated with the evolution of some resistance gene clusters via analysis of the synteny among Arachis, Lotus, and Medicago 59 . Several other hypotheses such as duplication, gene conversion, and unequal crossing-over have been proposed to elucidate R gene cluster and evolution 1 38 . Therefore, further studies will be needed to better understand common bean R gene evolution.

In this study, 365 PvRGLs were identified from 454-derived sequences and common bean ESTs in GenBank using data-mining. As a result, 105 and 237 PvRGLs were mapped to the P. vulgaris FPC physical map and genetic map, respectively. RGL-tagged markers were developed for 25 unique PvRGLs, including 11 STS and 19 CAPS markers. These methods and results will help develop more RGL-tagged molecular markers that can be used further for genetic mapping and the isolation of resistance genes from common bean in the future.

ZL participated in conceiving the research, sequence analysis, developing molecular markers, PvRGL mapping, and writing the manuscript. MC participated in material preparation, developing molecular markers, and contributed to writing of the manuscript. AT participated in PvRGL sequencing and contributed to the writing of the manuscript. VK participated in conceiving the research, providing the 454 sequences, and contributed to writing of the manuscript. All authors read and approved the final manuscript.