Despite the fact that sequencing costs have reduced significantly in the past few years and may continue to do so in the future, short-read shotgun data without annotation does not provide the information needed for analysis of genetic and linkage data and its functional implications. Moreover, development of polymorphic markers and mapping of millions of contigs generated by next generation sequencing is prohibitively expensive despite the development of high-throughput genotyping platforms. Data from animal systems indicate that survey sequencing (1-2X genome coverage) combined with high resolution RH maps deliver a comprehensive map of over 10,000 loci in which over 90% of the markers were ordered . When applied to plants, this combined approach can provide a dense physical map that may be used for crop improvement and selection, map-based gene cloning, and comparative analysis of various species. However, RH mapping relies heavily on the availability of an adequate RH population rich in detectable deletions. Wheat is an allohexaploid species, meaning that it contains three sets of homoeoloci for each gene and deletions in one or more of these loci can be potentially buffered by the others. Viable plants carrying large deletions can be obtained as demonstrated by the production of the wheat deletion bin genetic stocks . The AL8/78-DGRH1 is a whole genome panel designed for the RH mapping of all seven D-genome chromosomes of Ae. tauschii.
Radiation dose for development of RH panel
The selection of an appropriate dose for irradiation is an important step in the development of radiation hybrids in plants. Different genotypes may respond differently to radiation treatments. The hexaploid wheat cultivar Chinese Spring showed a survival rate of 21% when irradiated with 350 Gy of γ-rays  compared to 70% survival observed for the synthetic hexaploid wheat SW58 irradiated at the same dose. It was also observed that the seeds of LDN showed significantly higher survival compared to another durum wheat cultivar (both tetraploid) when irradiated with 350 Gy (unpublished data). Since SW58 contains the AB genome from LDN , a better response of SW58 and LDN to γ-radiation suggests that LDN genome may be more tolerant to radiation damage. Although the biological basis of irradiation tolerance in these lines is not known, the results of the present study suggests that for any RH panel, it is important to optimize the radiation dosage for the particular genotype.
The selection of an appropriate irradiation dose is also crucial as previous studies have observed an increased marker loss and mapping resolution with the increase in dose of γ-irradiation [18, 22, 24]. However, it is impossible with the RH approach in plants to go beyond a certain level of radiation as the increase in dosage leads to decrease in seed germination, plant survival, fertility and poor vigor [14, 18, 24]. A large decrease in plant survival above 350 Gy and lack of plant recovery from the 550 Gy irradiated seeds was observed in this study. This is likely due to the fact that the irradiated cells go through mitotic cell division and highly fragmented nuclei are not likely to survive the selection during plant development.
Based on the findings, the 450 Gy was the best possible dose for AL8/78-DGRH1 panel development. Additionally, as the higher doses of radiation cause higher number of breaks and rearrangements [18, 22], a larger number of markers will be needed to develop linkage groups . The 350 Gy RH lines were developed to complement the 450 Gy RH lines as the combination of RH panels of different resolution can yield a better contiguous map .
Effect of crossing scheme and role of genome imprinting
The methodology presented here to develop a RH panel for the D-genome of Ae. tauschii involves two interspecific hybridizations. First, the synthetic hexaploid wheat SW58 (AABBDD) was developed by crossing tetraploid durum wheat LDN (AABB) to A. tauschii (DD) accession AL8/78 . Secondly, the irradiated SW58 was crossed to LDN and pentaploid DGRH1 seeds were obtained, which allow for an exclusive identification of deletions in the D-genome of Ae. tauschii.
The F1 seeds developed from the interspecific crosses mostly show low germination and poor vigor that is attributed to possible crossing barriers. The most common post-zygotic reason for failure of wide hybridization is an embryo abortion due to poor endosperm development. The AL8/78-DGRH1 seeds developed from the reciprocal crosses (LDN × SW58 and SW58 × LDN) showed varying degree of endosperm development. The endosperm development was poor in most seeds when irradiated SW58 was used as male. The endosperm appeared more developed when SW58 was used as a female. This could be a reason for the differences observed in germination of AL8/78-DGRH1 seeds belonging to reciprocal crosses (~3% vs. ~80%). This difference could be attributed to genomic imprinting in endosperm resulting from differences in the ploidy levels of the two parental genotypes [51–53]. Endosperm is well developed whenever it is diploid [AAABBBDD in case of SW58 (♀) × LDN (♂)] for the whole set of extra seven chromosomes (DD in our case) and is wrinkled when it is haploid for the D chromosomes [AAABBBD in case of LDN (♀) × SW58(♂)] [54, 55]. Although there are chances that homologous chromosomes of the D-genome have deletions, it is highly unlikely that both have the same deletion, therefore, endosperm containing a pair of D-genome chromosomes even with deletions has a better chance of development. The differences in endosperm may also be due to genomic imprinting which is the result of monoallelic gene expression in a parent-of-origin–dependent manner . In Arabidopsis embryogenesis, there is a strong evidence of maternal epigenetic pathways controlling the parental contributions in plant embryos . Additionally, in cases of interspecific crosses, the correct balance between maternal and paternal genomes is disturbed as in the cross between durum and synthetic hexaploid wheat, and crosses between species with different ploidy levels, leading to significant differences in the success of reciprocal crosses .
GECN and its implication on RH panel development
When irradiating seeds, the diploid germline cells of the fully developed embryo are the radiation target. So, if the GECN is greater than one, the plant germinating from the M1 seed (RH0 seed in case of RHs) will be chimeric because independent mutations occur in each of the cells. The GECN has been reported to vary between different species. A GECN of two in Arabidopsis thaliana, Nicotiana plumbaginifolia[59–61], and soybean , four in flax , four or more in corn [47, 64], and six in barley  have been reported. To our knowledge, this is the first time that the GECN is calculated from a RH1 population and for wheat. The information about GECN of the species under study is important in any RH or mutation-based experiments, as it can be used to estimate the rate of recovered deletions or other types of mutations. The probability of recovering a RH line with a deletion at a given locus in any RH family (or a mutant in M2 family) depends upon the number of plants screened from that family, the GECN for that particular plant species, and on the frequency of the mutation/deletion. The probability increases with the number of plants screened and decreases as the GECN increases. The AL8/78-DGRH1 panel represented 312 RH1 families, with an average of 4.8 siblings per family. With a GECN = 5 and screening of ~5 lines per RH1 family, approximately 40% of the deletions available in this panel for the 35 SSR marker loci were identified. The GECN of 5 indicates that at least 30 seeds in an RH1 family need to be analyzed to recover the same mutation at 95% confidence. Therefore, the GECN information for wheat is important in selecting informative lines from an RH1 population or identifying mutants in an M2 population. In wheat RH mapping, GECN of ≥5 offers an additional advantage of identifying multiple RH lines with different deletion patterns from the same RH1 family developed by a single cross. The 399 RH1 lines showing deletion(s) for one or more of the 35 SSR markers belonged to 211 crosses and showed up to five unique genotypes per cross. This indicates that more informative lines can be obtained from few crosses, thus, reducing the efforts required for developing a RH panel.
Distribution of marker deletions in RHs
It has been estimated that less than 1% of recombination occurs in ~30% of the proximal regions of wheat chromosomes . This means that map-based cloning of genes in the recombination-poor regions is not feasible utilizing the traditional recombination based approach. The uneven distribution of recombination also leads to variation in genetic to physical distances across the chromosome. RH mapping uses retention/loss frequency as a result of breaks caused by radiations, which are expected to be random. Therefore, the mapping power of RHs is expected to be uniform across the length of a given chromosome.
The five markers selected from each chromosome mostly represented different deletion bins and were genetically far from one another. The mapping data of these markers shows that the marker loss frequency within a chromosome is homogeneous for all chromosomes except for chromosome 1D and 7D. For 1D the heterogeneity was due to two markers showing higher retention. Similar results have also been reported earlier for the markers present in the same deletion bin (1DS5-0.70-1.00 and 1DL2-0.41-1.00) . This could be due to the presence of critical gene(s) essential for the plant survival in this region leading to comparable lower observed marker loss than other chromosomal regions.
The distribution of deletions along a chromosome can be viewed from the five markers of chromosome 4D which represented different physical positions including centromeric bins of both long (C-4DL9-0.31) and short arm (C-4DS1-0.53) as well as sub-telomeric (4DS1-0.53-0.67) and telomeric region (4DL12-0.71-1.00). No preferential loss of markers for any region was observed. Similar results have been reported in maize , wheat , and cotton . These results suggest that a uniform mapping resolution across the genome can be achieved with RHs, including the low recombination centromeric regions, which is not possible with genetic mapping. However, studies based on larger number of randomly distributed molecular markers of different types will provide a more detailed picture of the distribution of deletions along the length of the chromosome, and will help to determine, whether or not, the radiation-induced breakage is effected by chromosome structure or any other biological phenomenon.
Selected RH lines for developing a high-resolution physical map of D-genome
The success of RH mapping depends on marker loss/retention frequency in a panel. Simulation studies have suggested that a retention/loss frequency of 50% would be optimal . However, most RH mapping studies conducted in animals showed average retention frequency of about 20-30% . The marker loss observed in this population was much lower than what was observed in animals, but was comparable to that earlier reported in wheat . In plants, it is not possible to obtain the hybrids with marker loss comparable to human or animal RH panels as viable and fertile plants from the irradiated seeds need to be recovered.
The informative panel of 399 lines selected in this study has an average marker loss about five times higher (9.9%) than the whole unselected DGRH1 panel (2.1%). The effectiveness of this selection strategy was tested by screening a subset of 92 random RH1 lines with 60 additional markers. However, a high correlation (r=0.94) was observed between deletion frequency of the RH lines based on screening with a small (35) and large set (95) of markers. Moreover, almost all the lines (>92%) detected with multiple chromosome (>3) breaks using the larger set were also included in the set of informative lines identified using only 35 markers. This shows that the RH lines with deletions identified using few (35) markers have multiple deletions across the genome. Thus, the strategy to select highly informative lines using few markers from across the whole genome is as effective as selection based on almost three times larger marker set. However, selection with a larger set of markers detected additional informative lines, suggesting that the these lines possess smaller and unique deletions, which could be extremely useful when the purpose is to fine map a particular genomic region.
Combination of different marker systems for a better contiguous map
Different marker systems represent different types of DNA sequences and potentially different distributions in the genome. Three different types of markers (RJM, SSR, EST) were used to characterize a subset of AL8/78-DGRH1 panel. The marker loss for RJMs was much higher and identified almost three times more lines with deletions as compared to SSRs or ESTs. A simple explanation would be that ESTs and SSRs mainly represent gene rich regions , potentially more prone to cause loss in vitality when deleted as compared to RJMs which usually span the inter-genic repetitive space . Any deletion in the gene clusters will reduce the chances of survival of the particular RH line, leading to the detection of fewer lines with deletions among the surviving plants. Comparatively, plants with deletions in the repetitive DNA have more chances of survival. However, repetitive DNA is present in both genic and inter-genic regions [2, 67] which could be a possible explanation for the wide range of marker loss (1.1-19.5%) observed for RJMs. The other explanation could be found in the specific marker design. RJMs are designed to span unique retrotransposon insertion sites. These insertion sites are typically characterized by a chromatin state more susceptible to DNA breakage . Recent findings indicate that radiation damage and the subsequent DNA repair might also prefer these exposed chromatin regions to cause breakage . This would also explain why RJMs have the tendency to be deleted with higher frequency. Similarly, the large amount of variation in chromatin state in intergenic regions would explain the wide range of marker loss for RJMs.
The abundance and uniform distribution of retrotransposons across the genome makes RJMs  ideal candidates for high density RH mapping. SSR markers, on the other hand, have been used extensively in wheat for the construction of genetic maps and the identification of marker-trait associations , http://wheat.pw.usda.gov. Therefore, SSRs can serve as excellent anchor points for mapping anonymous markers and as a linking bridge between the available genetic information and the high density RH maps. Gene-based markers like ESTs are an efficient tool for inter-species comparative genome analysis [15, 38].
The use of different type of markers to characterize a subset of AL8/78-DGRH1 panel showed that deletions in the majority of the lines (80%) were detected by either of the three marker type. Similarly, a recent study in wheat reporting the first physical map of a wheat chromosome, which used different classes of marker types to anchor BAC contigs, observed that ~ 70% of the contigs were anchored with only one marker type . These observations clearly suggest that different marker classes have different advantages for genomic research and also cover different regions of the genome. Therefore, a combination of various markers is better suited to obtain a complete RH/physical map for whole genome assembly.
Combining RH and genetic mapping: A better approach for genome mapping
Dependence of genetic mapping on recombination makes it difficult to conduct map-based cloning of genes in the recombination-poor regions. The requirement of polymorphism for genetic mapping also makes it more difficult to map gene-based markers  which are invaluable tools for comparative studies across species. The selection pressure tends to remove any non-advantageous allelic polymorphisms in the gene space making gene-based markers highly monomorphic among individuals of the same species. In wheat, 9.9% polymorphism was observed for EST-derived SSRs compared to 35.5% polymorphism observed in case of genomic SSRs .
The presence/absence aspect of RH mapping which do not require genetic polymorphism may help in drastically increasing the number of mapped markers including gene-based markers. From the 14 ESTs mapped on the chromosome 2D using 92 random RH1 lines (only 16 were critical lines), only six could be mapped in a diploid bi-parental population of 572 F2 plants , suggesting that RH mapping has the potential to increase the number of mapped markers using comparatively fewer lines. The 2D-RH map also provided 5.2 fold higher resolution (cM/cR ratio of 1:5.2) than the genetic map , confirming that RH mapping indeed provides higher resolution than genetic mapping. At the same time, the availability of hundreds of genetic maps and genetically mapped loci in wheat  is an asset and this information may provide better consensus maps. In addition, the genetic maps could easily be used for identifying QTL/genes for any phenotypic trait segregating in that population. Therefore, RH mapping has the potential to complement a genetic mapping approach by mapping the markers in low recombination regions and making it possible to map a larger amount of monomorphic gene-based markers.
Radiation hybrids for BAC contigs assembly
The IWGSC has established a clear road map to achieve the complete sequencing of the wheat genome . This includes the creation of high quality physical maps for each of the 21 chromosomes, individually. In this direction, the first physical map of a single chromosome was reported for 3B . The authors described the assembly of 1,036 contigs with an average size of 0.78 Mb, covering ~82% of the 3B chromosome. Although 1,443 molecular markers were developed from the BAC contigs, not all were polymorphic or could be uniquely mapped into a marker scaffold despite using information from 13 genetic mapping populations, especially for those regions that have limited recombination . Apparently the use of genetic maps alone might not be enough to generate sufficient resolution for the IWGSC initiatives  due to non-uniform distribution of recombinants along the length of the chromosome and low level of polymorphism in bi-parental genetic populations. In humans [71, 72] and animals [73–77], RH mapping played a major role in the contig assembly leading to whole genome sequencing .
Considering the success of RH approach in human and animal genome assembly, the AL8/78-DGRH1 panel was made to help the assembly of BAC contigs from the wheat D-genome project. Although an effort similar in scale to the one described here is not available as reference for the plant kingdom, preliminary studies suggested that a map resolution of ~200 kb is achievable with RH in plants [8, 23, 25]. Similar level of resolution was achieved in the AL8/78-DGRH1 panel as indicated by analysis with a set of markers for which the physical distance was known.
Past studies in plants have reported the detection of up to thirteen deletions for a single chromosome in an individual RH1 line [14, 18, 24, 25]. However, as observed in the present study and also reported earlier  the number of detectable deletions may increase by genotyping the panel with more markers. Assuming an average of 3 random deletions for a particular chromosome and BAC contigs of a size observed by Paux et al.  for 3B (482 kb), the selected set of 399 highly informative AL8/78-DGRH1 lines is adequate to anchor and order all the BAC contigs of any D-genome chromosome.
The resources generated here provide a foundation for the development of next generation, high-density RH maps incorporating thousands of marker including SSRs, RJMs, ESTs etc. These maps will provide a marker scaffold to develop complete physical maps leading to the final sequence assembly of the seven wheat D-genome chromosomes.
Potential applications of RHs for functional genomics
Cytogenetic stocks in wheat have been used extensively to study genes/phenotypes associated with a particular chromosome or its segment [78, 79]. However, due to the availability of limited number of these lines, this approach offers an even lower resolution than genetic mapping. The association of phenotypic and genotypic data, both recorded on RH1 lines developed during the present study, may help assign the phenotype to genomic deletions as is done in case of genetic mapping or deletion bin mapping [78, 79], but at much higher resolution. This has been successfully illustrated by mapping the genes in low recombination regions in wheat [14, 80].
The first-generation radiation hybrids (RH1s) contain D-genome chromosomes with rearrangements in a monosomic condition. Variation is expected in the transmission of D-genome chromosomes to subsequent generations [81, 82]. It also means that it will not be possible to replicate the phenotypic data recorded on the DGRH1 lines, unless the segregating lines are confirmed for the presence of a particular chromosome with the phenotypes. However, as an alternative RH maps can be used to identify the RH1 lines showing deletions for the markers associated with any particular trait of interest. Using this information, the parental M2 families of specific RH1 plants may be used to select homozygous and stable deletion mutants for the flanking markers (and the markers in between) of the gene/QTL. A set of such overlapping deletion lines and their trait data will help in physically placing the gene in a small segment of the chromosome. In a similar fashion, this approach may also be used for understanding the phenotypes associated with known genes (reverse genetics) by screening the DNA of the whole set of AL8/78-DGRH1 lines generated during the present study, for deletions of a particular gene sequence and then studying their parental M2s.