INTRODUCTION

Top Abstract Introduction RESULTS DISCUSSION METHODS References

Tandem repeats represent an important proportion of vertebrate genomes and have been classified as satellites, midisatellites, minisatellites, and microsatellites according to the overall length of the entire array. In higher vertebrates, (CA)_n microsatellites are the most numerous, with an average distance between two microsatellites of ~25 kb (Stallings et al. 1991). Ninety percent of human (CA)_n microsatellite arrays are <40 bp and <1%-2% are longer than 30 repeats (Weber 1990). Minisatellites repeat units are usually 10-100 nucleotides long, and the array spans 0.5-100 kb. Chromosomal distribution of minisatellites in the human genome is highly skewed towards telomeres and ancestrally telomeric regions (Amarger et al. 1998).

The initial classification of minisatellites and microsatellites has now been strengthened on biological grounds by the demonstration that different modes of evolution operate on these two types of structures. Microsatellites mutate by replication slippage processes because of mispairing between the two strands during replication. They are stabilized by variant repeats, whose presence facilitates detection of the slipped strand DNA by the mismatch repair system (Strand et al. 1993; Heale and Petes 1995). Minisatellites mutate predominantly in the germ line (Jeffreys and Neumann 1997) through mechanisms, including gene conversion-like events, presumably arising from DNA double-strand breaks (DSBs), insensitive to internal variations within the tandem array (Buard and Vergnaud 1994; Jeffreys et al. 1994). However, a number of intermediate situations raise the question of the border between the mini- and microsatellite classes. For instance, mutation rates at some minisatellites including MS1 (D1S7) are sensitive to mismatch repair deficiencies (Hoff-Olsen et al. 1995) reminiscent of a microsatellite behavior. At the other end of the spectrum, some human (CA)_n repeats have extremely long alleles, with internal heterogeneity (Wilkie and Higgs 1992). Also, the origin of both classes of tandem repeats is still poorly understood. Microsatellite arrays may arise by replication errors or as a result of nonhomologous end-joining repair following DNA DSB events (Liang et al. 1998), which can create de novo (CA)_n > 20 stretches. Unequal crossing-over or replication slippage between fortuitous short-direct repeats have been invoked to provide the initial duplication event of some minisatellites in human and yeast (Haber and Louis 1998).

To better understand the nature and origin of surprisingly long human (CA)s, we developed a technology that efficiently identifies clones containing long CA-rich sequences by a simple hybridization procedure. This approach was applied to human cosmid and PAC genomic libraries. The analysis of a subset of sequences strongly supports the conclusion that most long human CA-rich sequences are imperfect. The genome distribution of positive clones is highly skewed towards telomeres and minisatellites can usually be found in the vicinity. This observation is further strengthened by the parallel investigation of the currently available chromosome 7 human sequence data. Twenty-two new minisatellites have here been successfully identified, establishing the validity of this approach of minisatellite cloning by vicinity with long (CA)s.

	RESULTS

Top Abstract Introduction RESULTS DISCUSSION METHODS References

Identification of Probes Appropriate for the Identification of Long CA Arrays

CHROMOSOME 1 COSMID LIBRARY SCREENING AND SEQUENCE ANALYSIS OF SOME POSITIVE CLONES

CHROMOSOMAL ASSIGNMENT OF POSITIVE COSMID CLONES

View larger version (22K): [in this window] [in a new window]   Figure 1   Chromosomal assignment of the clones detected in the human genome after hybridization of R62 probe on cosmid and PAC libraries. PAC () or cosmid () localized by FISH and linkage; PAC () or cosmid () localized by FISH or linkage. The two semicircles represent one cosmid clone revealing two locations by in situ hybridization.

Application of the Methodology to the Screening of a Total Human Genome PAC Library

Probe R62 detects clones with a very good signal-to-background ratio and will not detect a (CA)₂₂ array [but would still detect a longer (CA)₄₀ array, independently characterized from a pig cosmid library; data not shown]. R62 was thus selected to hybridize a high-density filter carrying ~20,000 independent PAC clones, corresponding to one human genome equivalent. The 42 clones giving the strongest signal were successfully assigned to a chromosomal band by FISH and/or linkage analysis and are represented by squares on the 550-band karyotype presented Figure 1. Twenty-five PACs are assigned to a terminal band (60%), 4 are juxtacentromeric (9.5%), and 13 are interstitial (30.5%).

Identification of Minisatellites Within Positive PAC and Cosmid Clones

The cosmid and PAC clones identified by R62 screening were searched for minisatellites as described in Amarger et al. (1998). The DNA from each clone was digested separately with three combinations of two restriction enzymes: AluI and HaeIII; AluI and HinfI; HaeIII and HinfI. Seventy-three cosmid fragments with a size above 1.3 kb after the double digestion were excised from agarose and tested for the presence of a minisatellite by hybridization on a Southern blot. Three minisatellites (CEB117, CEB118, CEB119) were isolated (Table 2). Using the same approach, 316 PAC fragments were tested. Eighteen minisatellites derived from 15 independent PAC clones were identified. Their main characteristics (allele size, polymorphism) are presented in Table 2. Twenty out of the 22 new minisatellites identified are derived from the telomeric PAC or cosmid clones. One (or more) minisatellite was identified in half of the telomeric PAC clones. PAC 1 contains (at least) four minisatellites: UPS17, UPS21, UPS22 (Table 2), in addition to CEB 70, which was characterized previously and independently (Spurr et al. 1994). PAC 50 contains two minisatellites: UPS6 and UPS7.

Parallel Investigation of Sequence Databases

The current status of publicly available human sequence data is reflected at http://www.ncbi.nlm.nih.gov/genome/seq/. Significant progress has already been achieved for a number of chromosomes, such as chromosomes 7, 17, 21, and 22, so that the screening of genome libraries can be compared to some extent to the direct screening of genome sequence. We selected chromosome 7 for further investigations, because the available sequence is relatively well distributed along the whole chromosome (i.e., in contrast with chromosome 17) and because the distribution of minisatellites on chromosome 7 has been well documented in earlier reports (Amarger et al. 1998). At the time of this investigation, ~54 Mb of sequence data was available, corresponding to 30% of a total estimate of 170 Mb for chromosome 7. Figure 2 presents some of the results obtained by searching and locating minisatellites and long CA sequences along the chromosome. As a reminder, Figure 2A (left) is compiled from this report and Amarger et al. (1998) and locates minisatellite loci obtained by screening cosmid or PAC libraries. Figure 2B presents the density of tandem repeats with a repeat unit of 20 nucleotides or more, spanning at least 1000 nucleotides as identified in the sequence data using the tandem repeat finder described in Benson (1999). Figure 2C presents the relative density of long (spanning at least 300 nucleotides) CA-rich sequences detected by a FASTA search against the chromosome 7 sequence data using a 800 bp-long (CA)₄₀₀ as the query. None of the matches in this range is a perfect CA repeat. Four matches span >800 bp, two of which originate from 7q36 (no higher order organization of the degenerate CA rich array could be found; data not shown).

View larger version (25K): [in this window] [in a new window]   Figure 2   Comparison of genomic libraries and sequence database investigations for HSA Chr 7. (A) The position of minisatellite loci characterized in this report (one locus) and in Amarger et al. (1998) for chromosome 7, showing the predominantly, but not exclusively, telomeric location of minisatellites on this chromosome and the imbalance between the two chromosome ends. (B) The density of tandem repeats as detected in the sequence data (repeat unit >20 bp, repetition spanning at least 1000 nucleotides). (C) The density of long CA-rich sequences (in which the similarity, as detected by FASTA, with the 800-bp long CA400 query spans 300 bp or more). (D) The distribution of ESTs characterized so far along the chromosome. The numbering (1-11) refers to the eleven bins defined in the Methods section. The largest peak, in B and C, corresponds to a density of, respectively, two and three qualifying objects per megabase of sequence available for this bin.

	DISCUSSION

Top Abstract Introduction RESULTS DISCUSSION METHODS References

Minisatellites and microsatellites are two important classes of tandem repeats used as genome markers. Minisatellites have been shown to be useful tools to detect chromosomal rearrangements in a number of pathological situations, including mental retardation (Flint et al. 1995; Giraudeau et al. 1997). Some of them are suspected to be involved in gene regulation (Bennett et al. 1995). Very unstable human minisatellites have been characterized (Jeffreys et al. 1988; Vergnaud et al. 1991). The mutation rate is apparently increased by environmental agents such as radiation (Dubrova et al. 1997). To characterize new human minisatellites, as well as to investigate the boundary between mini- and microsatellites, we have devised a strategy enabling the rapid cloning of long CA-rich sequences from total genomic libraries by hybridization screening. Five CA-rich probes were evaluated for their ability to discriminate long CA-rich sequences from ordinary microsatellites. The long perfect (CA)_n array does not discriminate against fragments containing a (CA)₂₀ array. The imperfect synthetic tandem repeat 16C46 [16-bp repeat unit containing an internal stretch of four (CA)s], is more appropriate but cross-hybridizes with a cosmid clone containing a (CA)₂₂ array (data not shown) and fails to detect many CA-rich sequences (Table 1). Probe R85, with internal stretches of up to three (CA)s, detects clones with weak intensity. Probe R62 appears to be a very good compromise. It will not detect a perfect (CA)₂₂, although it will detect a longer (CA)₄₀ stretch. It detects many clones, with a good signal-to-background ratio, as compared to the synthetic 14C32 array. In contrast with 14C32, R62 also detects complex stretches of imperfect (CA)s devoid of higher-order organization (Table 1).

First, to reveal the chromosomal distribution pattern of long CA sequences, a chromosome 1-specific cosmid library was screened. Among the 22 cosmids studied, 13 originate from chromosome 1 (Fig. 1). The nine cosmids coming from other chromosomes, including five telomeric loci, presumably reflect some contamination of the library. The fact that the actual contamination of the library is less than the proportion of non-chromosome 1 cosmids in our selection (40%) suggests a telomeric bias in the contamination of the chromosome 1 library. Seven of the 13 chromosome 1 cosmids are within the terminal 1p36.3 band, where a minisatellite cluster was previously and independently described (Amarger et al. 1998). Two are localized on 1p34.35 where minisatellite MS1 (D1S7) is localized. Two others are localized in a juxtacentromeric region where the MUC1 gene characterized by tandem repeats units has been isolated. Another is localized in 1q42, containing minisatellite MS32 (D1S8). Overall, the chromosome 1 distribution pattern of long CA-rich sequences is highly similar to the chromosome 1 minisatellite distribution pattern that is shown in Amarger et al. (1998) or that can be deduced from the NIH/CEPH Collaborative Mapping Group (1992) data, suggesting that the procedure could be applied on a larger scale for the identification of minisatellite associated regions.

For this purpose, a whole-genome PAC library was screened using the R62 probe to enable the cloning of new minisatellite sequences in the vicinity of CA-rich sequences. A significant proportion (25/42) of the R62-positive PAC clones studied are assigned to a terminal band of the karyotype and 13 of them contain new minisatellites. The juxtacentromeric location of four (9.5%) PAC clones may reflect a peculiar behavior of these regions or may indicate an ancestrally telomeric location. Thirteen clones (30.5%) are interstitial. One of them is assigned to band 2q13 (Fig. 1) which is the position of a well-characterized chromosome fusion site (IJdo et al. 1991). Another one is located on 1p31-p32, where one minisatellite was described previously (Amarger et al. 1998). This human chromosomal region is homologous with 6qter in pig (Amarger et al. 1998). As shown here in PAC 1 and PAC 50 (Table 2), the use of large insert clones further emphasizes the clustering of minisatellites within telomeric regions (Vergnaud et al. 1993).

The predominantly telomeric distribution is highly reminiscent of the distribution of minisatellites across the human genome and clearly different from the even distribution of microsatellites. In good agreement with this, a similar result is obtained by the investigation of sequence databases using a FASTA search (Pearson and Lipman 1988) and (CA)₄₀₀ as the query. Although long perfect (CA)s (>200 bp, e.g., accession no. Z81056 from Caenorhabditis elegans) are represented in the database, none of these originate from primates or even other mammals (data not shown). Figure 2C shows the density of hits spanning at least 300 bp, across human chromosome 7. All such hits are imperfect, CA-rich stretches, with or without higher-order redundancy. The distribution is almost identical to the patterns shown in Figure 2, A and B, which reflect chromosome 7 minisatellite distribution. No obvious correlation is seen between the chromosome 7 gene density presented in Figure 2D and the minisatellite and long (CA)s distribution, with the exception perhaps of segment 7, band 7q22, which is a common peak (Fig. 2).

	METHODS

Top Abstract Introduction RESULTS DISCUSSION METHODS References

High-Density Filters from Human Genome Libraries

High-density filters corresponding to a human chromosome 1 cosmid library were obtained from the Max-Planck Institute for Molecular Genetics. This library is represented by two high-density filters with 20,000 clones spotted on each membrane. Each clone is named by the number (c112) of the library and a specific number.

High-density filters corresponding to a human PAC library were obtained from the Roswell Park Cancer Institute (RPCI) center (http://bacpac.med.buffalo.edu/; the RPCI6 segment was used).

Perfect and Scrambled (CA)_n Arrays Probes

A perfect long (CA)_n probe and the two imperfect (CA)_n arrays 14C32 (GACACACTCACAGC)_n and 16C46 (CACACACATGCACATA)_n were synthesized as described in Vergnaud (1989). 14C32 and 16C46 were designed so as to contain a maximum of three and four uninterrupted CA repeats, respectively. The natural scrambled (CA)_n arrays R62 (EMBL accession no. AC AJ000072) and R85 (EMBL accession no. AJ000073) were selected among rat minisatellite sequences (Pravenec et al. 1996; Amarger et al. 1998). R62 and R85 repeat units are (CACACT)_1-2CACAGYRR (14 or 20 bp) and (CAGGACA)_1-2 GTGARCACA (16 or 23 bp), respectively.

Probe Labeling and Hybridization

The DNA fragments were recovered from agarose by centrifugation through glass wool as described by Heery et al. (1990). The probes were labeled with [-³²P]dCTP Institute of Chemical and Nuclear (ICN) by the random priming procedure (Feinberg and Vogelstein 1984). Hybridization was done as described in Vergnaud (1989) in an hybridization oven. After hybridization, the filters were washed in 1× SSC/0.1% SDS or 0.1× SSC/0.1% SDS. Hybridization and washing were done at 60°C (screening of library filters) or 65°C (hybridization of Southern blots).

Subcloning and Sequencing

Restriction digest fragments were recovered from agarose using the Jetsorb kit (Bioprobe System). The fragments were ligated into SmaI Puc 18 vector (Pharmacia) before transfer to Escherichia coli XL1 strain (Stratagene) by electroporation.

Recombinant plasmids were sequenced using ³³P-labeled direct and reverse M13 primers with the Delta Taq sequencing kit (U.S. Biochemical) in a Perkin Elmer GenAmp PCR System 9600 thermocycler.

Identification of Minisatellites Within PAC and Cosmid Clones

DNA from each PAC or cosmid clone was digested by AluI and HaeIII, AluI and HinfI, or HaeIII and HinfI. Fragments >1.3 kb in size were recovered from agarose and hybridized to Southern blots carrying two reference individuals digested separately by AluI, HaeIII, HinfI, and PvuII, as described in Amarger et al. (1998).

Chromosomal Assignment by Linkage Analysis

Linkage analysis was performed on the CEPH (Centre d'Etudes du Polymorphisme Humain) panel of human families. Genotypes were managed using GENBASE, developed by Jean Marc Sebaoun (Spurr et al. 1994). Linkage files output were converted to CRIMAP file format using the LINK2CRI utility software written by John Attwood. CRIMAP version 2.4 (Green et al. 1990) was used for the analyses.

Chromosomal Assignment by FISH

Cosmid or PAC DNAs were labeled with biotin by nick translation. After overnight hybridization on target chromosome spreads, slides were washed in 2× SSC at 37°C. Probes were detected with FITC-avidin and analyzed with an epifluorescence microscope (DMRB-Leica) equipped with a CCD camera driven by the Powergene system from Perceptive Scientific International (PSI).

Sequence Database Searches

Chromosome 7 sequence data (54 Mb available at the time of this investigation) were retrieved from the National Center for Biotechnology Information (NCBI) site at (http://www.ncbi.nlm.nih.gov/genome/seq/). Tandem repeats were identified using the online software accessible at http://c3.biomath.mssm.edu/trf.html (Benson 1999). Large CA-rich sequences were detected using FASTA (Pearson and Lipman 1988) and a (CA)₄₀₀ synthetic sequence as the query. The FASTA analysis was done using the computing facilities provided by Infobiogen (information at http://www.infobiogen.fr/). Each sequence contig was assigned to a bin along the chromosome. Eleven bins of equal size were defined. The horizontal bars presented in Figure 2 represent the density of the object category per megabase of sequence in the corresponding bin. The current distribution of human ESTs on chromosome 7 was retrieved from the NCBI site (http://www.ncbi.nlm.nih.gov/genemap/).