Directed Gap Closure in Large-Scale Sequencing Projects

RESULTS

For the selective isolation of gap sequences, we used two DNA populations. The tester was prepared from X. fastidiosagenomic DNA that was digested with Sau3AI or BamHI, respectively, and ligated to PCR adapters for later amplification. For the driver, the whole cosmid library was used after digestion with the same restriction enzymes. A 2000-fold excess of driver was used in the subtractive hybridization experiments. The subsequent PCR resulted in difference products enriched for fragments, which originated from the gap areas.

Gel electrophoresis of the difference products showed a smear ranging from 100 to 900 bp for the Sau3AI-digested DNA, and discrete bands between 500 bp and ∼2 kb for the BamHI-digested DNA (Fig. 1) as one would expect, because the enzymes have 4-bp and 6-bp recognition sites, respectively. A total of 314 clones from both the Sau3AI and the BamHI preparation were picked and end-sequenced. Taking into account redundant information, the 314 sequences could be reduced to 168 sequence clusters, of which 159 contained X. fastidiosa DNA. Of these, 141 clusters contained only one or two sequences, whereas 27 clusters consisted of 3 to 26 sequences. The latter resulted mainly from the experiment with BamHI. The nine remaining clusters without X. fastidiosa sequence were of unknown origin and showed only low homology to any known sequence stored in databases.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 1.

Gel separation of the difference products. S and B indicate the enzymes (Sau3AI and BamHI) used for the initial restriction digest. (M) 100-bp marker ladder (MBI Fermentas, Germany).

Of clones that represent the 159 X. fastidiosa clusters, 123 (74% of the total) had an insert DNA which is not contained in the cosmid library but is part of the final X. fastidiosa genome sequence. A detailed comparison showed that with the exception of two tiny gaps of ∼0.7 and 1.3 kb, all other gaps were represented by sequences in our gap-filling library (Fig.2). Moreover, the 1285-bp long X. fastidiosa plasmid (pXF1.3), which is not present in the cosmid library, was represented by three fragments covering more than half of the plasmid's sequence. The end sequencing provided more than 28 kb of gap-specific sequence information, with individual read lengths of up to 700 bp. In reality, the clones contained more gap sequence, of course, that was not accessed, however, because only end-sequencing was performed. Moreover, further clone picking and sequencing would have produced significantly more gap-specific sequences, because the redundancy observed in the library resulting from the Sau3AI experiment was quite low.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 2.

Representativity of gap-filling sequences. The lower row of blocks represents position and length of the gaps in the Xylella fastidiosa cosmid library as determined elsewhere (Frohme et al. 2000). Bars in the top row indicate the location and distribution of 123 gap-specific fragments within the 2700-kb X. fastidiosa genome.

DISCUSSION

The approach presented here provides the means to produce, selectively, sequence information for an accelerated gap closure in genomic sequencing projects. The experimental procedures involved are simple to perform and provide a high proportion of gap-specific sequences. Restriction enzymes with both 4- and 6-bp recognition sequences were found to be useful. Usually, the fragments resulting from a digest with the former can be sequenced completely by a single read. The overall larger fragments produced by the latter enzyme might be a more efficient choice for the closure of wider gaps, because the sequence redundancy could be reduced by using them as templates in a primer walking strategy. In any case, however, the resulting fragments do not only provide sequence information but are also helpful starting points for other approaches of gap closure. The sequence information is useful for the generation of PCR or sequencing primers located within gaps, for example, or for the generation of probes to be hybridized to other library types to identify potentially gap-spanning clones. In this last respect, a labeled difference product may be used directly as a probe for screening another library.

The relatively high portion of false positives — one quarter of the clones did not contain gap-specific Xylella DNA — is probably the result of the single step of subtraction enrichment, done in an attempt to reach a compromise between specificity and representation. Further enrichment steps, as are usually performed in conventional RDA experiments, would probably have resulted in a smaller number of unwanted non-gap sequences due to a stronger selectivity. However, it could have caused a bias toward fewer fragments, resulting in a less complex DNA population.

As in all techniques that enrich DNA fragments by subtraction, external contamination could be a problem in this gap-filling procedure. However, the enrichment factor is considerably lower in this setup than in conventional RDA. Therefore, contamination is less of a problem. Nevertheless, all precautions routinely adopted for highly sensitive, PCR-based protocols were applied, particularly during the preparation of the starting material for the tester. In addition to potential contamination, we experienced another technical problem: Chimeric clones can occur, if a low ratio of adaptor:DNA is used during the ligation at the very start of tester preparation. Thus, a careful determination of the DNA concentrations is critical.

Although the results presented here are based on a genomic cosmid library, the procedure should be applicable irrespective of the type of vector system that is being used. At an earlier stage of the project, only a subset of (then) sequenced cosmids had been used in a preliminary experiment. As expected, a more complex fragment pattern could be seen in the gel electrophoresis. These difference products, however, were not analyzed in more detail, because the sequencing had proceeded further. In shotgun sequencing projects, plasmid libraries form the basis of analysis. Because the clonability is usually better for shorter inserts, one would expect fewer real gaps in such projects. Still, the procedure should be useful nevertheless, even if only for a verification of the completeness of the clone coverage.