Report

Whole-Genome Shotgun Optical Mapping of Deinococcus radiodurans

See allHide authors and affiliations

Science  03 Sep 1999:
Vol. 285, Issue 5433, pp. 1558-1562
DOI: 10.1126/science.285.5433.1558

Abstract

A whole-genome restriction map of Deinococcus radiodurans, a radiation-resistant bacterium able to survive up to 15,000 grays of ionizing radiation, was constructed without using DNA libraries, the polymerase chain reaction, or electrophoresis. Very large, randomly sheared, genomic DNA fragments were used to construct maps from individual DNA molecules that were assembled into two circular overlapping maps (2.6 and 0.415 megabases), without gaps. A third smaller chromosome (176 kilobases) was identified and characterized. Aberrant nonlinear DNA structures that may define chromosome structure and organization, as well as intermediates in DNA repair, were directly visualized by optical mapping techniques after γ irradiation.

Detailed, structural knowledge of whole microbial genomes is of primary importance to many genomic studies, but this information has been difficult to obtain. Pulsed-field gel electrophoresis (PFGE) plus Southern (DNA) blot analysis (1) provides primary genome information but does not confidently size large circular genomes and frequently obscures the analysis of large episomal elements. Although an eight-enzyme restriction map of the Escherichia coli K12 genome was constructed in 1987 by Kohara et al. (2), this required a laborious approach involving partial digestion of 3400 phage clones followed by Southern blot analysis. Physical maps of theSaccharomyces cerevisiae genome were also prepared by painstaking restriction mapping of clones (3). Modern microbial genome analysis uses shotgun sequencing, followed by finishing efforts (4, 5). Whole-genome restriction maps may become an indispensable resource for large-scale genome sequencing projects. They facilitate sequence assembly by providing a scaffold for high-resolution alignment and verification of sequence assemblies (contigs), accurate genome sizing, and discernment of extrachromosomal elements (6).

Optical mapping is a system for the construction of ordered restriction maps from individual DNA molecules (7, 8) and has been used to prepare restriction maps of a number of clone types, including phage clones (9), yeast artificial chromosomes (10), bacterial artificial chromosomes (6), and, more recently, an entire electrophoretically separated chromosome (∼1 Mb) from Plasmodium falciparum (11). An optical mapping approach for whole bacterial genome analysis is feasible because we can now mount and map extremely large, randomly sheared DNA molecules (0.4 to 2.4 Mb) that are digested with high cutting efficiency (70 to 90%). These parameters critically control the success rate of assembling the fragments and are well modeled by prior probabilistic (Bayesian) analysis (12). The contigs covering a whole genome were initially assembled manually, or later with the Gentig algorithm (13), which automatically computes contigs of genomic maps.

To efficiently collect such large molecules, we developed a semiautomated image acquisition system that collects successive images and correctly assembles them into one large superimage while maintaining proper pixel registration between images. A new image analysis system was developed [Visionade (14)] that enables markup of molecular images, allows for editing, and automatically calculates fragment masses and cutting efficiencies. A λ bacteriophage DNA sizing standard was used as follows. First, the total integrated fluorescence intensity of the standard was determined to be in the correct range. Second, the size of each genomic fragment in a particular image was calculated by dividing the fluorescence intensity of the DNA fragment by the average fluorescence intensity of the standards in the image and then multiplying this by the size of the standard. The images were not used as data if the cutting efficiency along the length of the molecule was less than 75%.

The development of Gentig enabled the rapid assembly of raw maps into a complete genome-wide map in minutes rather than months, with negligible false positives. Contigs of the E. coli genome were assembled with Gentig into a consensus map, which both reproduced the map constructed by hand and correlated with the map predicted by sequence. Gentig automatically generates contigs from optical mapping data by repeatedly combining the two islands that produce the greatest increase in probability density, excluding any contigs whose false positive overlap probability is unacceptable. The standard deviation, digestion rate, false cut rate, and false match rate are given as parameters to Gentig and affect the false negative probability and hence the number and distribution of contigs and gaps. Gentig also facilitates assembly of whole-genome maps for much larger eukaryotic genomes such as Plasmodium falciparum (15). Here, Gentig assembled a consensus map derived from 50 molecules and found 147 out of the 150 cut sites predicted from the manually assembled map. The three cuts that Gentig failed to call were due to the lower depth of coverage applied in this benchmarking exercise. Whereas manually constructed maps of E. coli and D. radioduransrequired several months to completely assemble, Gentig required only 20 to 30 min to generate contigs.

The radiation resistance of D. radiodurans may be a serendipitous result of its ability to survive periods of severe dehydration, which also fragments DNA. Efforts to understand the detailed mechanisms underlying this uniquely effective feat of DNA repair are now centered on the annotation results of a complete genome sequence. We thought an optical map would facilitate ongoing sequencing efforts, as a scaffold for sequence assembly, and might identify aberrant DNA structures associated with mechanisms of DNA repair. Before mapping the uncharacterized D. radiodurans genome, we benchmarked our genomic mapping system by mapping the sequenced (4) E. coli K12 substrain MG1665 (16). There was a <1% size difference in the optical versus sequence map [the same error rate was then seen when a Nhe I optical map of D. radiodurans was compared with preliminary sequence information (17)].Deinococcus radiodurans R1 (American Type Culture Collection 13939) is a radiation-resistant organism with a comparably sized genome (18). A representative molecule is shown in Fig. 1A. The final map was assembled without gaps at an average depth of 35×, using 157 molecules, with an average fragment size of 29 kb (19). This depth of coverage gave the map very low errors (20). Contigs were also assembled with Gentig (Fig. 1B) (21). The genome size was calculated to be 2.61 Mb by manual assembly and 2.59 Mb by Gentig assembly. Gentig assembled a consensus map derived from 100 molecules. Manual editing can assign more molecules to the contig in a way that is not yet modeled in Gentig. To further confirm our map, we constructed a rare-cutter, Not I, map (macro) and aligned it, using Gentig, with the Nhe I (micro) map, which was analyzed by Southern blots (22).

Figure 1

Optical mapping of D. radioduransgenomic DNA with Nhe I. (A) Representative DNA molecule, 0.7 mm (2.4 Mb) in length, spanning six microscope fields, used to make an optical map. (B) Large chromosome restriction map generated by shotgun optical mapping. The outer circle shows the consensus map; the inner circles show the contig from which the consensus map was generated. The contig was assembled from 100 molecules by means of the Gentig algorithm. Nhe I fragment sizes (in kilobases) can be measured from the figure (20). Colors are arbitrarily assigned to homologous overlapping fragments.

Large extrachromosomal elements are notoriously difficult to characterize. For example, circular DNA molecules (25 kb) identified inEntamoeba histolytica (23) migrated as high molecular weight bands on pulsed-field gels, resembling chromosomes, and were only characterized after single-molecule analysis with electron microscopy. Optical mapping of the D. radioduransgenome also provided insight into the genomic structure of the organism. The genome is reported to be 3.17 Mb in size, as assessed by PFGE (24). We calculated the genome to be 2.6 Mb in size by optical means and also discovered the presence of a large episome, 415 kb in size, that was previously considered to be a part of a large single chromosome and which accounted for the missing portion (Fig. 2). We believe this episome to be a second (mini) chromosome, on the basis of its size and 1:1 stoichiometry with the large chromosome as analyzed by PFGE. This second chromosome has been shown from sequence assembly to contain genic regions critical to cell function (25), and thus is not likely to have resulted from a duplication event. In addition, we found no map homology between the second chromosome and any other genomic region. Interplasmidic and intrachromosomal recombination have been found to occur in D. radiodurans at high frequency after exposure to ionizing radiation (26). Dessication may induce similar changes. Consequently, the same type of chromosomal architecture may not exist in all D. radiodurans strains. We also saw images containing a third circular molecule, which was sized at 176 kb [SD = 10 kb, n = 8 (25)]. Nhe I cut sites were apparent, but an insufficient number of molecules were collected to enable us to make Nhe I maps. Other smaller chromosomes or large plasmids (27) may be present in the D. radiodurans genome, although we did not map any such molecules.

Figure 2

Second (mini) chromosome of D. radiodurans identified by optical mapping. The image shows a representative second chromosome mapped with Nhe I (fragment sizes for the consensus map are 29.8, 4.6, 26.7, 35.4, 11.8, 145.0, 6.3, 69.8, and 85.5 kb). Second chromosomes were also mapped with Not I (fragment sizes for the consensus map are 11.5, 72.3, and 331.2 kb). Colors are arbitrary and correspond to fragment sizes, starting at the 12 o'clock position.

Whereas the genomes of most organisms are irreversibly shattered by the effects of high levels of ionizing radiation, D. radiodurans is able to efficiently reconstruct an intact genome through means that are not yet fully understood. Dissecting its ability to deal with severe DNA damage may uncover new general mechanisms of DNA repair. The postulated DNA repair system of D. radioduransis in part facilitated by multiple chromosome copies held in register. According to this model, double-stranded breaks are repaired by recA-dependent recombination, with the undamaged DNA duplex acting as the template (28, 29). One model for holding chromosome copies together in sequence alignment is by the presence of persistent Holliday junctions (30, 31). We saw no evidence of DNA molecules containing Holliday junctions, which might have been visualized as bundles of DNA molecules or molecules with a χ-shape. We also saw no evidence of back-to-back repeated regions containing Nhe I sites. These would have been noticeable as multiple ordered fragments of the same size. We did, however, see large repeated DNA molecules, which shared the Not I digestion pattern of the second chromosome (32). The pattern (10, 70, 340 kb) was repeated up to three times, and 10 such molecules were observed. We do not know the origin of such molecules, but we speculate that these may be observable replication intermediates.

Images of DNA extracted from cells after γ irradiation of 17.5 Gy (Fig. 3) were examined from both wild-type and recA-deficient [rec 30 (30)] strains. We have previously shown that DNA fragments flanked by identical 4-kb sequences are efficiently circularized after γ irradiation in both wild-type and recA strains (31). Deinococcus radiodurans contains hundreds of short repeats (∼150 base pairs) and numerous long repeats, providing ample substrate for circularization of large portions of the genome. Such circles could be expected to protect vulnerable DNA ends produced by double-strand breaks from exonucleolytic degradation. Furthermore, homologous recombination among these circles could progressively generate larger circles, ultimately restoring the circular chromosomes without rearrangements. Optical mapping did not show the presence of circles of any dimension, although it should be noted that the resolution of light microscopy would not enable us to characterize circles smaller than ∼30 kb.

Figure 3

DNA repair after irradiation, visualized by optical mapping. DNA samples were extracted at different time points from D. radiodurans R1 (wild type) cells. The irradiation and recovery conditions (time interval after irradiation) for each image were: (A) 0 Gy, 0 hours; (B) 17.5 Gy, 0 hours; (C) 17.5 Gy, 2 hours; (D) 17.5 Gy, 4 hours; (E) 17.5 Gy, 6 hours; (F) 17.5 Gy, 16 hours; (G) 17.5 Gy, 24 hours. Scale bar, 20 μm. Histogram analysis (34) showed consistent elongation of DNA molecules into full-length genomic molecules. The recA-deficient strain (rec 30) showed no evidence of repair, and the observable number of molecules decreased with time as a result of cell death (35).

However, the imaging results graphically showed the repair of the genome in wild-type cells, but not in recA cells, by unbranched linear elongation of extracted molecules from small kilobase fragments to essentially intact chromosomes. Although this is consistent with previous conclusions based on gel electrophoresis and sucrose gradient technologies, this finding eliminates circular intermediates as a generalized mechanism of DNA repair. What protects the ends of these molecules from degradation remains unknown. After irradiation we estimated that the average fragment size was ∼15 kb (Fig. 3B), corresponding to about 200 double-strand breaks per genome.

The complete restriction mapping of whole genomes may catalyze the development of new modes of genome analysis, unhindered by the need to generate and map large numbers of clones. Because ensembles of single molecules are analyzed, small amounts of starting material are required, enabling mapping of microorganisms that are problematic to culture. Shotgun mapping obviates the need for library construction with its associated cloning artifacts, and also enables mapping of organisms with DNA, which is difficult to clone. Perhaps most important, optical mapping renders new biological insight by readily providing a picture of the basic organization of an entire genome, revealing the number of chromosomes and the presence of extrachromosomal elements as well as providing a means of directly examining dynamic processes such as DNA repair. If maps can be rapidly generated, microbial populations can be analyzed at the whole-genome level to reveal genotypic differences that can be linked to phenotype.

  • * Present address: Genome Therapeutics Corp., 100 Beaver Street, Waltham, MA 02154, USA.

  • Present address: CNS Disorders, Wyeth-Ayerst Research, Princeton, NJ 08543, USA.

  • Present address: Smith Kline Beecham Pharmaceuticals, King of Prussia, PA 19406, USA.

  • § To whom correspondence should be addressed. Present address: Departments of Chemistry and Genetics, UW Biotechnology Center, University of Wisconsin, 425 Henry Mall, Madison, WI 53706, USA. E-mail: schwad01{at}mcrcr.med.nyu.edu

REFERENCES AND NOTES

View Abstract

Navigate This Article