INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

In the course of retrovirus infection, a DNA copy of the viral RNA genome is synthesized, and that DNA is then permanently inserted into the genome of the host cell. The integration of the viral DNA into the host genome is a crucial step in the life cycle: it is important for the efficient expression of progeny virus, and it is responsible for the ability of these viruses to persist and cause disease (36). If most or all of the regions of the host genome are accessible to retroviral integration (14, 36, 46), the frequency of use of potential sites varies considerably. This integration specificity correlates with local DNA structural features, which govern the accessibility of specific regions (37, 50).

Indeed, integration may be favored near DNase I-hypersensitive sites (32, 33) or CpG islands (23), suggesting a preference for transcriptionally active regions (26, 37). Human immunodeficiency virus type 1 (HIV-1) integration preferentially occurs near Alu elements (41) or topoisomerase cleavage sites (19). Recently, centromeric alphoid repeats were found to be selectively absent at HIV-1 integration sites (6). A nonrandom, compartmentalized integration in GC-rich isochores has been previously described for human T-cell leukemia virus type 1 (HTLV-1) (52), bovine leukemia virus (22), hepatitis B virus (51), and Rous sarcoma virus (34), while mouse mammary tumor virus has been found to integrate into GC-poor regions (35) of the host genome. At the nucleotide level, in vitro integration reactions and analysis of the flanking sequence of cloned viruses have shown that there is a preference for integration in A/T-rich regions (15, 17, 27, 39). A consensus sequence at the direct repeat flanking HIV-1 provirus genome has been shown by sequencing integration sites derived from cells infected in vitro (6) (41, 47).

The data summarized above result from in vitro studies, and only a few flanking sequences derived from naturally infected samples have been described to date. HTLV-1 is the causative agent of adult T-cell leukemia/lymphoma (ATLL), an aggressive T-cell malignancy and of tropical spastic paraparesis/HTLV-1 associated myelopathy (TSP/HAM), a chronic progressive neuromyelopathy. A study based on the analysis of four HTLV-1 integration sites derived from uncultured ATLL cells has suggested that the regions flanking the provirus were A/T-rich with a nucleotide composition bias in the 6-bp direct repeat generated by integration (13). In a previous work based on the analysis of 24 distinct HTLV-1 integration sites derived from asymptomatic carriers, the mean A/T content of the hexameric repeat was found to be 59% (49).

An inventory of 218 distinct HTLV-1 integration sites was made from infected individuals. The virus was found to integrate in A/T-rich regions and a weakly conserved sequence was identified at in vivo integration sites. Database analysis revealed that these features were not associated with any preference for integration in transcriptionally active regions or in repeat elements of the host genome.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Samples studied. Samples from 29 HTLV-1-infected individuals were analyzed. These samples corresponded to 12 ATLL, 10 TSP/HAM, and 7 asymptomatic carriers. DNA was phenol-chloroform extracted and ethanol precipitated from peripheral blood mononuclear cells (PBMCs), cerebrospinal fluid (CSF), lymph nodes, or skin biopsies.

PCR. Our strategy is summarized in Fig. 1. DNA was amplified as described by ligation-mediated PCR (LMPCR) or inverse PCR (IPCR) (7-10, 49). Both of these two methods allow the amplification of the HTLV-1 3' extremities, together with their cellular flanking sequences. For LMPCR, DNA was digested with NlaIII in 1× NlaIII buffer for 3 h at 37°C. Digestion was controlled by gel electrophoresis. DNA was phenol-chloroform extracted and ethanol precipitated. Digested DNA was ligated with BIO1 primer (49) by using T4 DNA ligase. This was followed by two phenol-chloroform extractions and precipitation. Ligated DNA was amplified for 100 cycles with the BIO2 primer alone (49). Conditions were 1× Stoffel DNA polymerase buffer, 1.5 mM MgCl₂, 50 pmol BIO2, a 150 µM concentration of each deoxynucleoside triphosphate (dNTP), and 10 U of Stoffel fragment of Taq DNA polymerase (Perkin-Elmer Cetus) in a final volume of 85 µl. First, 25 µl of a 1× PCR buffer containing dNTPs and primers were first loaded into a 750-µl tube, and an Ampliwax PCR Gem 100 (Cetus) was added to each tube. After wax layer formation by incubation at 75°C 10 min and cooling at room temperature for 15 min, 60 µl of the remaining reagent and ligated products were loaded. Thermal cycling parameters were as follows: 1 cycle of 94°C for 10 min and 100 cycles of 95°C for 45 s, 60°C for 45 s, and 72°C for 2 min, followed by a final elongation step of 10 min at 72°C. Ten microliters of this linear PCR reaction was used in a classical PCR amplification with the BIO3 and BIO4 primer pair (49). Amplification conditions were as before, with 40 pmol of each primer and 2.5 U of Taq polymerase all in a final volume of 100 µl. Thermal cycling parameters were as follows: 1 cycle of 94°C for 10 min and 35 cycles of 95°C for 45 s, 58°C for 45 s, and 72°C for 1 min, followed by a final elongation step of 10 min at 72°C.

View larger version (15K): [in this window] [in a new window]   FIG. 1.   Strategy to construct HTLV-1 flanking sequences and control sequences libraries. (A) Construction of HTLV-1 integration sites library. HTLV-1-infected DNA was amplified by LMPCR or IPCR. Amplified products were cloned and sequenced as detailed in Materials and Methods. (B) A ~1,500-bp segment spanning the 3' extremity of the provirus was generated by PCR amplification of a HTLV-1 provirus cloned in p4.39 plasmid (10). High-molecular-weight uninfected human DNA was first sonicated or digested with PvuII and EcoRV which cut blunt ends. Sonicated or digested DNA was then ligated with the 3' HTLV-1 purified construct as detailed in Materials and Methods. Then, LMPCR, IPCR, cloning, and sequencing were performed in the same conditions as with the infected samples.

Cloning and sequencing. PCR products were phosphorylated by T4 polynucleotide kinase and ligated with SmaI-digested and dephosphorylated M13mp18 replicative-form DNA as described earlier (49). After transformation of Escherichia coli XL1 by electroporation, recombinant M13 plaques were transferred in situ to nitrocellulose filters and screened by hybridization with the HTLV-1 long terminal repeat (LTR)-specific ³²P-labeled oligonucleotide BIO5 (49). Filters were first prehybridized at 42°C for 2 h in 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-1× Denhardt's solution-10 µg of denatured salmon sperm DNA per ml. Hybridization was carried out in a fresh solution at 42°C, and filters were washed in 2× SSC-0.1% sodium dodecyl sulfate at 42°C. Positive plaques were picked and prepared for DNA sequencing. Single-stranded templates were sequenced by using fluorescent dideoxynucleotides. The products were resolved on an Applied Biosystems 377A DNA sequencer with 377 software.

Control experiments. A control was used in order to compare the nucleotide composition of HTLV-1 flanking sequences with that of the uninfected DNA. As shown in Fig. 1B, a 1,494-bp segment spanning the 3' extremity of the provirus was first generated by PCR. This was performed by using an integration site-specific primer, PBSI (10), and an LTR-specific primer, MH2 (5'-CCCGCCAATCACTCATACAACC-3') that encompassed the 3' extremity of the HTLV-1 provirus cloned in the plasmid p4.39 (10). Amplification conditions and thermal cycling parameters were as follows: 1 cycle of 95°C for 10 min and 35 cycles of 95°C for 1 min, 58°C for 45 s, and 72°C for 2 min, followed by a final elongation step of 10 min at 72°C. Amplified products were purified and treated with the Klenow fragment of the DNA polymerase. As shown in Fig. 1, high-molecular-weight uninfected human DNA that derived from an uninfected blood donor was blunt-end digested with PvuII and EcoRV or sonicated. These three different methods were used in order to avoid any bias in the selection of DNA sequence during fragmentation. Digested or sonicated DNA was then ligated with the purified 3' HTLV-1 end. Then, LMPCR, IPCR, cloning, and sequencing were performed under the same conditions as with samples that were derived from infected individuals.

Comparison of integration sites with database sequences. Sequences with integration sites longer than 25 bp were analyzed by comparison to the nonredundant human sequence (nr) database (2), the human cDNA (dbEST) database (3), and the MONTH (January 1999) database (42) by using BLASTN with Search Launcher (1), FASTA (28), and Repeat Masker (See A. F. A. Smit and P. Green, RepeatMasker, at http://ftp.genome.washington.edu/cgi-bin/RepeatMasker) (6). Default parameters were used. The 25-bp cutoff was arbitrary chosen in order to avoid nonspecific homologies that could result from the alignment of shorter sequences with those of databases. A total of 18,118 bp of human DNA corresponding to 178 integration sites that were 25 bp were analyzed. For the 44 control sequences of 25 bp, 2,991 bp were analyzed. These 21,109 bp corresponded to 96% of the 22,018 bp we have sequenced. The lengths of flanking human DNA sequences analyzed ranged from 25 to 350 bp. The lengths of control sequences analyzed ranged from 25 to 150 bp. Similarities to repeated sequences were ranked in accordance with the Smith-Waterman parameter generated by RepeatMasker or by the probability of matching by chance generated by BLASTN (1) and by FASTA (29).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Construction of control sequences and HTLV-1 integration sites libraries. The goal of the study was to analyze HTLV-1 integration sites isolated by LMPCR or IPCR. Since PCR and cloning are influenced by target sequence composition and sequence size, a control was performed in order to assess the nucleotide composition of the uninfected human genome by using the same experimental procedure as that used in the analysis of HTLV-1 integration sites (see Materials and Methods and Fig. 1). This was done by sequencing flanking sequences of an HTLV-1 construct ligated in uninfected DNA. The three different methods used in the control DNA fragmentation process (PvuII digestion, EcoRV digestion, and sonication) ruled out any bias in the nucleotide composition of the control sequences. HTLV-1 flanking sequences isolated from infected individuals were subsequently compared to that of the control library.

HTLV-1 flanking sequences are A/T-rich. The A/T content was 51% (median, 50%; standard deviation [SD], 11%; range, 30 to 80%) after control experiments, a value significantly lower than that initially described for the human DNA (58%) (43). DNA digestion, amplification, and cloning may have accounted for the selection of such sequences. Based on these results, we performed a comparative analysis of HTLV-1 integration sites and control sequences.

4

Nonhomogenous distribution of the nucleotide composition within and without the hexameric repeat. Figure 2 represents the distribution of A/T nucleotides along the 40 first positions of both the HTLV-1 integration sites and control sequences. Deletion of the 3' LTR sequence was performed prior to sequence analysis. In addition, deletion of the first 31 bases of the p4.39 integration site, together with the 3 bases that correspond to PvuII and EcoRV restriction sites, was also performed when corresponding control sequences were studied. For the box plot analysis, points more than 1.5 times the interquartile range from the ends of the box were labeled as outliers. Points more than 3 times the interquartile range from the ends of the box were labeled as extreme values (see legend of Fig. 2).

View larger version (25K): [in this window] [in a new window]   FIG. 2.   Distribution of A/T nucleotides at each of the first 40 positions of the flanking sequences. The 218 HTLV-1 integration sites () and the 60 control sequences () were aligned, as described in the text, relative to the first base of the flanking sequences 3' to the provirus. The standard deviation of the A/T content is shown at each position of the HTLV-1 flanking sequences. The insert shows the distribution of the AT content along the flanking sequences. The boxes represent the difference between the 75th percentile and the 25th percentile of the AT distribution (i.e., the interquartile range). Within the inset box, the median is represented by a thick horizontal line. Lines from the ends of the box extend as far as the most extreme values not considered outliers. Points more than 1.5 times the interquartile range from the ends of the box are labeled as outliers () or as extreme values (*).

4

4

View larger version (25K): [in this window] [in a new window]   FIG. 3.   Nucleotide composition of the HTLV-1 flanking sequences. Sequences isolated after cloning LMPCR and IPCR products were aligned relative to the first base of the hexameric repeat sequences flanking the provirus.

Identification of the host sequences targeted for HTLV-1 integration in vivo. The matches to known sequences are summarized in Table 1, while Table 2 shows the distribution of HTLV-1 3' integration sites and control sequences with significant homologies to known nonrepetitive elements from databases. The sequences were classified as genomic noncoding nonrepetitive elements; most of these corresponded to GenBank high-throughput genomic sequences, transcription units, and repeat elements. Overall, the mean size of anonymous sequences was not significantly different than that of other sequences (100 versus 99 bp). As shown in Table 1, 84 (47%) of the 178 HTLV-1 sequences of 25 bp were found to match to known sequences. Of the 44 control sequences of 25 bp, 17 (39%) matched known sequences. For both libraries, the frequency of matches to repeat sequences was significantly higher than that for transcription units. Table 2 also shows that there was no significant difference in the distribution of matches between the HTLV-1 integration sites and the control sequences. In addition, there was no correlation between the clinical status accompanying to the DNA samples and the type of matches (not shown).

4

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In the infected cell, integrase functions as a component of the preintegration complex derived from the virus core. The main factors which may influence the choice for a chromosomal target site during infection include chromatin structure (26), target DNA sequence (15), host cell proteins, and virus-encoded proteins other than integrase (4). The present study was conducted in order to assess the selection of target sites during integration of the HTLV-1 provirus in vivo. To this end, HTLV-1 integration sites, isolated by IPCR or LMPCR amplification from naturally infected individuals, were sequenced. The nucleotide composition of the flanking sequences was analyzed, and their homologies to known sequences were identified. The nature of the HTLV-1 flanking sequences was compared to that of uninfected DNA. Since PCR and cloning are influenced by the structure and the nucleotide content of the target sequence, we designed a control experiment in order to isolate uninfected DNA sequences by using a strategy identical to that used for the analysis of HTLV-1 flanking sequences. The A/T content was found to be 51% after this control experiment, a value significantly lower than that initially described for human DNA (58%) (43). This indicates that the A/T content is underestimated by IPCR and LMPCR and emphasizes the need to perform a comparative analysis of HTLV-1 integration sites with sequences isolated by the same manner. Results from the comparative analysis show that there is a nucleotide composition bias at HTLV-1 integration sites. The virus integrates into A/T-rich regions with a nonhomogeneous distribution of residues 3' to the provirus. These characteristics do not appear to be associated with a preference for integration in transcription units or repeat elements.

Preference for A/T-rich regions as a common feature for in vivo and in vitro integration. HTLV-1 flanking sequences (~19 kb) isolated from naturally infected individuals were found to be significantly more A/T-rich than control sequences. A preference for A/T-rich regions has been previously described for insertion by transposons (36) or Ty elements (28) and for integration of adenovirus (24), spleen necrosis virus (40), avian myeloblastosis virus (16, 17) and Moloney murine leukemia virus (31). By hybridization of a viral probe with compositional fractions of HTLV-1 cultured cells DNA, HTLV-1 sequences have been found to be distributed in the 46 to 61% A/T range of the host genome (52). Recently, four HTLV-1 flanking sequences (183 bp) isolated from four ATLL patients were found to have an average A/T content of 63% (13). Altogether, these data indicate that a preference for A/T-rich regions characterizes both in vitro and in vivo integration.

The target DNA structure affects HTLV-1 integration site selection. Integration of proviral DNA into the host genome generates short direct repeats of 4 to 6 bp as a result of DNA repair to the cellular sequence flanking the integrated provirus (45). Analysis of the nucleotide sequence surrounding the integrated provirus indicates that the size of the direct repeat is characteristic of each virus. For MMLV, the middle two positions of the 4-bp direct repeat are preferentially occupied by AA, TT, or AT dinucleotides (31). Similarly, a preferential AT pairing toward the central portion of the direct repeat has been previously described in direct repeat generated by the insertion of yeast Ty elements (28) and in the target of site-specific recombination by FLP recombinase (44). For HIV, analysis of the 112 published direct repeats shows that there is a preference for a G residue at the first position, for T at the second position, for A at the third and fourth positions, and for C at the fifth position (6, 41, 47). Here, alignment of the 218 HTLV-1 clones showed a weak consensus sequence within the hexameric repeat flanking the HTLV-1 provirus. However, only the C/G content at position 1 of the hexameric repeat corresponded to an outlier as shown in Fig. 2. In fact, we have found a significant nucleotide composition bias at positions within and without the hexameric repeat. Indeed, there was a preference for C or G residues at position 1, for A residues at position 8, for C residues at position 9, and for A or T residues at position 16. These weakly conserved motifs observed in the vicinity of the provirus 3' end may constitute a propitious binding site for the preintegration complex. Alternatively, they might correspond to binding sites for other cellular or viral factors that interact favorably with the integration machinery.

Lack of evidence for favored HTLV-1 integration near known sequences. Nucleotide composition of the database sequences with significant homologies with HTLV-1 integration sites reflected the nature of the corresponding matches. Indeed, transcription unit sequences harbored the lowest A/T range, which correspond to DNA regions of high gene concentration. However, there was no significant difference in the frequency and distribution of matches to known sequences between HTLV-1 and control sequences. In contrast to results from a previous analysis of HIV-1 integration in SupT1 cells (6), we found that HTLV-1 can integrate into alphoid repeats in vivo. The frequency of HTLV-1 integration in repeat element was 33%, a value similar to that previously described for HIV integration in cultured cells (6). The frequency of matches to known transcription units was not significantly different between HTLV-1 and control sequences. This suggests that transcriptionally active regions are not preferred targets for HTLV-1 integration in vivo.

HTLV-1 integration may alter gene expression in vivo. In contrast to Ty3 elements (5, 11, 12), Ty1 elements (20), and retroviruses that induce chronic leukemia in animals (25), HTLV-1 integration does not preferentially occur near cellular sequences such as proto-oncogenes (38). However, about 6% of the HTLV-1 proviruses analyzed in the present work were found to be integrated into transcription units. This suggests that in some cells, HTLV-1 integration may alter gene expression in vivo. Given the elevated proviral load in HTLV-1-infected individuals, a substantial number of clonally expanded cells must have disrupted transcription units (6% of a large number is substantial). It is possible that some of these events could contribute further to clonal expansion.

Basis of HTLV-1 target site selection in vivo. HTLV-1 replicates mainly via the mitosis of its host cells (49) and the number of distinct circulating clones of infected CD4 T cells reflects the number of integration events. In a previous work, the overall number of such events has been estimated in an asymptomatic carrier in whom 20 distinct clones of infected cells were evidenced (10). Although this number is higher in TSP/HAM and in ATLL (7, 9), the present collection of HTLV-1 flanking sequences appears to correspond to a representative sample of in vivo HTLV-1 integration events.