Sequencing of long stretches of repetitive DNA

Repetitive DNA is widespread in eukaryotic genomes, in some cases making up more than 80% of the total. SSRs are a type of repetitive DNA formed by short motifs repeated in tandem arrays. In some species, SSRs may be organized into long stretches, usually associated with the constitutive heterochromatin. Variation in repeats can alter the expression of genes, and changes in the number of repeats have been linked to certain human diseases. Unfortunately, the molecular characterization of these repeats has been hampered by technical limitations related to cloning and sequencing. Indeed, most sequenced genomes contain gaps owing to repetitive DNA-related assembly difficulties. This paper reports an alternative method for sequencing of long stretches of repetitive DNA based on the combined use of 1) a linear vector to stabilize the cloning process, and 2) the use of exonuclease III for obtaining progressive deletions of SSR-rich fragments. This strategy allowed the sequencing of a fragment containing a stretch of 6.2 kb of continuous SSRs. To demonstrate that this procedure can sequence other kinds of repetitive DNA, it was used to examine a 4.5 kb fragment containing a cluster of 15 repeats of the 5S rRNA gene of barley.

of every chromosome in different taxa, and in some cases revealed the connection between polyploids and their potential diploid progenitors.
Our group is interested in understanding the molecular structure of SSRs in long stretches of repeats within the heterochromatin regions of the genomes of cereals. Repetitive DNA in heterochromatin can change rapidly in nucleotide sequence or copy number during evolution. However, this fraction of the genome is little understood in many species including barley, and it is considered essential for the epigenetic maintenance of centromeric function as well as for other cellular, developmental and evolutionary processes. In addition to exiting approaches, a strategy is here reported that allowed the sequencing of a 7.3 kb DNA fragment of the barley genome made up of SSRs motifs and could represent an alternative method for sequencing of long stretches of repetitive DNA. To demonstrate the value of this strategy, the sequence of a 4.5 kb fragment containing repeats of the tandem repetitive 5S rRNA gene of barley was sequenced. The method relies on the use of linear plasmids to clone fragments rich in SSRs motifs, and digestion with exonuclease in controlled serial deletions.

Results and Discussion
The screening of a BAC library containing the barley genome using oligo (ACT) 5 as a probe revealed several candidate clones rich in SSRs repeats (data not shown). To verify the presence of repeats, these clones were characterized by digesting them with different restriction enzymes which have no specific recognition sites within the repeat sequences analysed, allowing the identification of long SSR. In some cases the digestions showed smeared restriction signals in agarose gels, a previously described consequence of BAC clone instability 11 . The fragments obtained were hybridized with the same oligo (data not shown). For each digestion, several fragments of different size showed strong hybridization signals. A selection of fragments larger than 4 kb was used in subcloning and sequencing. Subcloning employing standard circular vectors was attempted, but the subclones obtained were highly unstable, hindering further experiments. A major drawback was the loss of repetitive DNA during the storing of the bacterial strain at − 80 °C. In many cases, plasmid DNA obtained from different bacterial cultures of the same stored subclone rendered differences in size; in some cases the entire repetitive region was lost. This kind of loss has been related to the instability of the cloned region caused by the formation of secondary structures among the tandem repeats in the clones 9 . In addition, the features of typical cloning vectors increase this instability 16 , e.g., the promoters present in the vector backbone might express cloned DNA causing the rearrangement or deletion of the repeated sequences.
To overcome these problems, fragments rich in SSRs were subcloned using a linear vector (Fig. 1A). This sort of vector has been shown useful in stably maintaining tandem repeat inserts of up to 30 kb 9 ; they also contain transcriptional terminators that prevent the expression of repeats. This enabled the cloning and manipulation of all fragments selected up to 20 kb in length (data not shown). The presence of repeats in these fragments was confirmed by the sequencing of the ends of the subclones using primers based on the pJAZZ vector (Fig. 1B). A subclone containing a fragment of around 7.5 kb rich in SSRs was selected to be entirely sequenced.
Attempts to sequence this fragment by the consecutive design of primers failed due to the inability to find unique specific sites for annealing. A strategy using exonuclease III was thus followed, based on previous work 17 . This relies on the production of consecutive, unidirectional deletions of the fragment by the exonuclease, and the posterior sequencing of these using vector primers. Exonuclease III removes nucleotides at the 3′ end of blunt or recessed ends and nicks although 3′ overhangs longer than 4 bases prevent the action of this enzyme 18 . The creation of 3′ overhangs at one end and 5′ protruding or blunt end at the other end of the fragment ensures deletion occurs in one direction only. For the present motif, a preliminary assay was performed to select the restriction enzymes that generated these kinds of end in the selected subclone. A protective ApaI unique restriction site was found in the vector 8 bp upstream of the fragment, and a SwaI blunt unique site located 87 bp downstream of the fragment (Fig. 1B).
After digestion of the subclone with both restriction enzymes, three aliquots of 5 μ g each were treated with exonuclease III over different time intervals to generate progressive deletions of the fragment. The first aliquot was treated for 2 min and the remainder for 90 s longer each time. The digestion rate of exonuclease III at 37 °C is 450 bp/min 19 , therefore consecutive deletions of some 700-900 bp must be expected in each aliquot. However, in our experience, a mixture of fragments with different deletions, ranging from a few bases to more than 2000 bp is obtained in each aliquot. Thus, the exonuclease III digestions of the three aliquots generated enough deletions of some 400 bp to cover the entire fragment (Fig. 1C). During the exonuclease treatment the vector is degraded because the covalently closed 'hairpin' ends are not protected 9 .
Cloning of the fragments with deletions into the linear vector is necessary to multiply the DNA of interest so there is sufficient to be sequenced, and to avoid the presence of post-exonuclease treatment remains of vector fragments. The treated aliquots were therefore electrophoresed in an agarose gel and the fragments purified and cloned (Fig. 1D). An alternative method was also tried, in which it was sought to purify entire fragments from the agarose gel after the digestion with the two restriction enzymes but before the treatment with exonuclease III. However, digestion failed. The latter enzyme is highly sensitive to NaCl (concentrations as low as 20 mM have been known to affect its activity 19 , and the buffers supplied in agarose purification kits contain high NaCl concentrations. Despite repeated washes with 70% ethanol, this NaCl may have stalled all exonuclease III activity. Several more procedures were tried too, but these yielded only very low DNA concentrations preventing any further experiments. Clones containing deleted fragments were analysed by the toothpick plasmid assay 20 . This allowed the size of the insert to be estimated directly in the bacterial colonies by gel electrophoresis without the need for clone purification. This method is very effective when many clones need to be analyzed. Moreover, the use of a linear vector facilitated the estimation of clone size compared to the use of standard circular plasmids. Given the large size of some clones (up to 20 kb), overnight electrophoresis in agarose gels was performed. This led to the identification of bacterial colonies containing cloned fragment deletions with size differences of around 400 bp. A set of clones covering the complete original fragment were selected and purified (Fig. 2). The sequencing of these clones was performed using primers designed on the vector arms. In each case, the primer next to the non-protected end was  (Fig. 1D). Since the fragment deletions can be inserted in either direction during the cloning procedure, tests must be performed in order to select the correct primer. In this work, the digestion of clones with SfiI produced a banding pattern that clearly identified the orientation of the cloned insert.
Sequencing of the selected clones provided a set of reads. To build the complete sequence of the original 7.5 kb fragment, reads were assembled following the order in which the fragment deletions were obtained (Fig. 1E). The reads from consecutive deletions with differences of around 400 bp were assembled (since Sanger sequencing usually produced reads of over 500 bp, the last 100 bp at least overlapped for consecutive reads). Assembly was facilitated by the presence of single base changes present in the overlapping region of consecutive reads (Fig. 3). Such nucleotide differences increase end-user confidence when matching up overlaps 8 . In the present work, although the sequencing of 20 clones containing consecutive deletions of 400 bp was sufficient to obtain the complete sequence of the 7.5 kb fragment, more than one clone was sequenced over the same part of the fragment in order to improve the reliability of the results. No differences were ever detected between clones of the same region, supporting the method's reliability. The reads were assembled using CodonCode Aligner software, but most had to be manually assembled since the program was unable to resolve ambiguities and to overlap reads in the correct order. Manual assembly of more than 40 reads of continuous trinucleotide repeats was easily performed if they were assembled in the same order in which deletions were produced. In all cases, a few single changes in the repeat motif was enough to unambiguously assemble the reads. The consensus sequence for the assembled reads was a fragment of 7333 bp, with a region of trinucleotide repeats of around 6.2 kb (Fig. 4). In the hypothetical case of perfect repeats (never observed in the present study), assembly can be correctly performed if the exact size of deletion is known. This could be done by analyzing the size of the deleted fragments using an automatic fragment analyzer.
To show that this procedure can also be used to sequence repeated DNA sequences other than SSRs, a fragment of 4.5 kb containing the 5S rRNA gene was sequenced. In barley, 5S rRNA is produced from clusters of repeated genes 21 -maybe thousands of copies -on chromosome arms 2HL, 3HL, 4HL and 7HS. Every copy contains a coding region of approximately 120 bp and a non-transcribed spacer (NTS) region of 100-700 bp 22 . This fragment was sequenced following the methodology described above, using ApaI and NotI enzymes to respectively generate the protected and non-protected ends in the fragment. The orientation of the cloned deletions was determined using the BssSI restriction enzyme. Reads were assembled in strict accordance with the order in which the deletions were produced, resulting in a consensus sequence of 4585 bp (Fig. 5) composed of 15 copies of the 5S rRNA gene.
Analysis of the molecular structure of repeated DNA has been hampered because of technical difficulties relating to the sequencing of repeats. In a pioneering attempt to analyze the molecular structure of SSRs, Ananiev et al. 11 sequenced fragments no longer than 700 bp using traditional cloning and sequencing strategies, although fragments of more than 2 kb rich in SSRs were detected in Southern blot experiments. Technologies developed for the sequencing of complete genomes have also failed because they incorrectly assemble the reads for repeated regions. Modern 3rd-generation sequencing technologies such as single-molecule real-time (SMRT) sequencing using the PacBio platform can obtain reads of up to 2.3 kb of 100% repeated DNA 3 , but the error rate is relatively high 23 . Further, this procedure depends on the number of SMRT cells used 24 . Nanopore sequencing technologies can produce reads of adequate length (tens of kb or longer), but the accuracy of the technique needs to be greatly improved 25 . An example of the difficulty of sequencing this kind of DNA is showed in the last release of the reference sequence of the Drosophila melanogaster genome 5 . After most of one decade, the highly repeat-rich regions,  including large satellite blocks and the ribosomal RNA genes and the centromere regions, are still inaccessible to current sequencing and assembly methods and remain poorly represented. For this reason, new sequencing technologies that produce very long reads of repeated DNA are required.
The proposed procedure provides an alternative method for sequencing of long stretches of repetitive DNA in addition to exiting approaches, and may be useful for the in-depth analysis of this type of DNA. The use of linear  vectors overcomes the problems associated with the stability of these sequences during cloning. In addition, the serial production of deletions with exonuclease III allows the correct assembly of reads, overcoming the problems derived from their overlapping. This method could be of use in the analysis of specific repeated DNA such as that found in centromere or telomere sequences or to characterize genes with a repetitive structure. The proposed method is currently being used to sequence long stretches of other SSRs.

Materials and Methods
Barley BAC library screening and clone characterization. A total of 115,200 clones of the HVVMRXALLhB BAC library of barley were screened at the French Plant Genomic Resources Centre (INRA-CNRVG) at Toulouse (France). Radioactively labelled oligo (ACT) 5 was used as a probe. Clones with strong signals were selected and their DNA purified using the NucleoBond Xtra BAC kit (Macherey-Nagel).
Southern blot experiments were conducted to characterize the clones. One microgram of DNA of each clone was digested with EcoRI, HindIII, BamHI, KpnI, BglII and SalI restriction enzymes. Fragments were electrophoresed on 0.8% agarose gels and transferred to a nylon membrane. Hybridization was performed using digoxigenin-labelled (ACT) 5 oligo as a probe. Positive fragments were subcloned into a SmaI site (blunt) on the pJAZZ linear vector (Lucigen) using the BigEasy v2.0 Linear Cloning Kits (Lucigen) following the manufacturer's instructions. End sequencing of each subclone was performed using the vector primers SL-1 and NZ provided with the kit.
Exonuclease digestion of subclones and sequencing. Two enzymes with unique restriction sites (one of them leaving a 4-base 3′ overhang) were selected to digest 30 μ g of DNA of the subclones, each cutting at different ends of the fragment. The digested DNA was then purified using the DNA Clean & Concentrator-5 kit (Zymo Research). Aliquots of 5 μ g were treated with 500 U of exonuclease III (Promega) at 37 °C for different lengths of time. The first was incubated for 2 min and then treated with 60 U of S1 Nuclease (Promega) for 30 min at room temperature to remove single-stranded tails. The reaction was then stopped by adding 1 μ l of stop solution (300 mM Tris base, 50 mM EDTA) and incubating at 70 °C for 10 min. The remaining aliquots were incubated with exonuclease over intervals increasing each time by 90 s, before being subjected to the rest of the treatment. The reaction products were run in 0.8% agarose gels and the digested fragments purified using the Speedtools PCR Clean Up kit (BioTools). Each fragment was end-repaired using the DNATerminator ® End Repair kit (Lucigen) and cloned into the pJAZZ vector as described above. Clones containing fragments were selected by the toothpick assay 20 , inoculating 1 ml of LB medium with appropriate antibiotics instead of the master plate. Selected clones were purified using the Zyppy Plasmid Miniprep kit (Zymo Research) and sequenced. Assembly of the sequences was performed using CodonCode Aligner software (CodonCode Corporation) with manual corrections.