INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Human T-cell leukemia virus type 1 (HTLV-1) is the causative agent of adult T-cell leukemia and HTLV-1-associated myelopathy/tropical spastic paraperesis (11, 31, 35, 49). HTLV-1 is the prototype of the HTLV-bovine leukemia virus family of complex retroviruses. It encodes a protein, Tax I, that enhances viral gene expression through sequences in the U3 region of the viral long terminal repeat (LTR) (7, 8, 40). Tax I can activate the viral promoter as well as a wide number of cellular promoters (2, 13, 15, 16, 23). It is generally believed that Tax I enhances transcription by interacting with cellular transcription factors that, in turn, bind the target DNA sequences.

The organization of the HTLV-1 LTR is illustrated in Fig. 1. The elements that impart Tax I responsiveness to the LTR consist of three imperfect 21-bp repeats (TRE-1s) (4, 9, 33, 36, 37, 39) and a region between the promoter-central and promoter-proximal 21-bp repeats (TRE-2) (24, 28). Each TRE-1 contains a near-consensus cyclic AMP response element (CRE) as well as GC-rich flanking sequences. At least two TRE-1s are required for activation by Tax (4, 24, 28, 36). Since Tax I does not efficiently bind TRE-1 DNA in vitro by itself (1, 12, 29), previous studies have focused on identifying cellular proteins that bind TRE-1 directly or in conjunction with Tax I. By a variety of in vitro techniques, these studies have identified CREB and other members of the CREB/ATF superfamily of transcription factors as the primary proteins that specifically interact with TRE-1 (17, 30, 42, 44, 50). All of the cellular proteins that bind the TRE-1 DNA or Tax I are bZIP proteins. It has also been shown that Tax I promotes dimerization of bZIP proteins and DNA binding (3, 34, 45).

View larger version (9K): [in this window] [in a new window]   FIG. 1.   Schematic of the HTLV-1 LTR and relative positions of oligonucleotides used for LMPCR. +1 signifies the start site of transcription. Relative positions of the HTLV-1-specific oligonucleotides used in LMPCR (see Materials and Methods for details) are also shown. Oligonucleotides ending in -S were used to analyze the lower strand, while those ending in -AS were used to analyze the upper strand.

The GC-rich flanking sequences are required for activation of each TRE-1 by Tax I both in vivo and in vitro (5, 10, 17, 28, 32, 51). It has recently been shown that Tax I can contact the DNA within the GC-rich flanking sequences in vitro (22). These contacts are at symmetric positions on either side of the CRE (19). Tax may act as an anchor for recruiting the cellular coactivator CREB-binding protein to the transcription complex.

The HTLV TRE-2 region, between the promoter-central and promoter-proximal TRE-1s, can also mediate transactivation by Tax I (24, 28). TRE-2 is not capable of Tax response by itself; it can impart a Tax response only in the presence of at least one TRE-1 (28). TRE-2 contains binding sites for a large number of transcription factors, including AP-2, HNF-3, Ets family members, NFB, and Sp1.

In this study, we have analyzed the in vivo protein occupancy of the HTLV-1 LTR in infected human T-cell lines by in vivo dimethylsulfate (DMS) footprinting. This is an important issue for HTLV-1 (as well as all other retroviruses), since actual regulation of the viral LTR occurs when the provirus is organized in the host cell chromatin. Although the previous in vitro studies were essential for dissecting the molecular mechanisms of Tax I activation, it has been unclear if all of the protein-DNA interactions predicted from the in vitro experiments occur in vivo in HTLV-1-infected cells. The in vivo footprinting described in this report addressed this question and provided interesting (and sometimes unexpected) answers.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Cell lines. In vivo footprinting was performed on the cell lines MT-4 (21) and MT-2 (26), both human T-cell cell lines isolated from cord blood lymphocytes that were cocultured with cells from patients with adult T-cell leukemia. The AIDS Research and Reference Reagent Program provided the MT-4 and MT-2 cells used in this study. The cell lines were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum.

Transfections and reporter assays. MT-2 or MT-4 cells were cotransfected with pHTLV-I LTR-SEAP (a reporter construct with the HTLV-1 LTR driving the secreted alkaline phosphatase reporter gene [SEAP] and either pRL-null (a promoterless Renilla luciferase reporter plasmid; Promega) or pRL-tk (a Renilla luciferase reporter plasmid driven by the herpes simplex virus thymidine kinase gene promoter; Promega). The ratio of reporter to coreporter DNA was 15:1. Transfections were carried out in triplicate using 4 µg of total DNA per transfection. The reagent used for the transfections was DMRIE-C (Gibco/BRL), and cells were transfected according to the manufacturer's instructions. The medium was sampled for SEAP activity 72 h posttransfection. The assays for SEAP were performed using a Phospha-Light chemiluminescent reporter gene assay kit (Tropix, Inc.) according to the manufacturer's instructions.

Western blots. Whole-cell protein lysates from MT-2 and MT-4 cells were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) on a 10% gel, and the proteins were transferred to a nitrocellulose membrane by using a semidry transfer unit. The membrane was then blocked in a solution of Tris-buffered saline containing 5% bovine serum albumin and 0.05% Tween 20 for 1 h at room temperature and incubated with a polyclonal antibody to Tax I (a kind gift from K.-T. Jeang, Bethesda, Md.) in blocking solution overnight at 4°C. The membrane was then washed repeatedly with a solution of 0.05% Tween 20 in Tris-buffered saline and incubated with a secondary antibody conjugated to horseradish peroxidase for 1 h at room temperature. The membrane was then washed as above, developed with reagents for chemiluminescent detection (Pierce), and exposed to film.

In vivo DMS treatment of HTLV-1-infected cells. Exponentially growing suspension cultures of MT-2 and MT-4 cells were counted and harvested by centrifugation at 500 × g. To obtain partial DMS methylation in vivo, three independent samples of 0.5 × 10⁸ to 1 × 10⁸ cells were treated with 1 ml of growth medium containing 1% DMS for 1 min at 37°C. Exposure to DMS was stopped by the addition of 49 ml of ice-cold phosphate-buffered saline (PBS) followed by immediate low-speed centrifugation. Residual DMS was removed by an additional PBS wash. The pelleted cells were resuspended in 0.3 ml of PBS. Genomic DNA was harvested by the addition of 2.7 ml of cell lysis solution (300 mM NaCl, 50 mM Tris-Cl [pH 8.0], 25 mM EDTA [pH 8.0], 0.2% [vol/vol] SDS, 0.2 mg of proteinase K/ml). As a control, genomic DNA was also harvested from untreated MT-4 and MT-2 cultures in parallel. The samples were incubated at 37°C for 4 h to overnight, after which the genomic DNA was phenol-chloroform extracted and ethanol precipitated.

In vitro DMS treatment of DNA. Genomic DNA from control cultures was subjected to DMS treatment in vitro by incubation with 1% DMS in H₂O for 1 min at 25°C. The reaction was stopped by the addition of ice-cold DMS stop buffer (1.5 M sodium acetate [pH 7.0], 1 M -mercaptoethanol, 100 µg of yeast tRNA/ml) immediately followed by the addition of 2.5 volumes of ethanol on dry ice. Samples were precipitated by incubation at 70°C for at least 30 min and pelleted by centrifugation in a microcentrifuge for 15 min at 4°C. DNA pellets were allowed to air dry for 10 min and resuspended in 200 µl of 1 M piperidine in H₂O for 15 min at room temperature prior to cleavage.

Piperidine cleavage. Following in vivo and in vitro DMS treatment, genomic DNA from each cell type was cleaved at all methylated guanine residues by incubation in 200 µl of 1 M piperidine for 30 min at 90°C. The piperidine was removed by lyophilization, and the cleaved DNA pellets were resuspended in 360 µl of TE buffer (10 mM Tris-Cl, 1 mM EDTA [pH 7.5]). Residual piperidine was removed by two successive ethanol precipitations (addition of 40 µl of 3 M sodium acetate followed by 1 ml of 100% ethanol) and incubation for 30 min at 70°C. DNA samples were pelleted by microcentrifugation for 15 min at 4°C and resuspended in 500 µl of TE buffer. The DNA pellets were ethanol precipitated a second time by the addition of 170 µl of 8 M ammonium acetate plus 670 µl of isopropanol and incubation for at least 30 min at 70°C. The precipitated samples were pelleted by centrifugation as above, washed with 500 µl of 75% ethanol, and centrifuged for 5 min at room temperature. The resulting DNA pellets were resuspended in double-distilled H₂O at a final concentration of 0.4 µg/µl.

LMPCR. Two micrograms of DMS-treated, piperidine-cleaved genomic DNA was used for ligation-mediated PCR (LMPCR) as described by Mueller and Wold (27), with minor modifications. Single-stranded DNA fragments with guanine residues at both termini result from the DMS treatment and piperidine cleavage. To provide appropriate substrates for linker ligation, double-stranded, blunt-ended molecules were generated by primer extension from an HTLV-1-specific oligonucleotide (Fig. 1, oligo 1). This first-strand primer extension was accomplished by incubation of 2 µg of DMS-treated, piperidine-cleaved DNA with 0.3 pmol of oligo 1, 1× Vent DNA polymerase buffer (New England Biolabs), 4 mM MgSO₄, 0.25 mM deoxynucleoside triphosphates (dNTPs), and 0.5 U of Vent DNA polymerase (New England Biolabs) in a total volume of 30 µl. The DNA was denatured at 95°C for 5 min, annealed by incubation at 55°C for 20 min, and extended by a subsequent incubation of 10 min at 72°C. Ligation of the unidirectional linker as described by Mueller and Wold (27) (Fig. 1) was completed by the addition of 20 µl of 110 mM Tris-Cl (pH 7.5)-17.5 mM MgCl₂-50 mM dithiothreitol and 25 µl of 10 mM MgCl₂-20 mM dithiothreitol-3 mM ATP (pH 7.0)-4 µM unidirectional linker (in 50 mM Tris-Cl [pH 7.7])-3 U of T4 DNA ligase (Gibco/BRL). This mixture was incubated at 17°C overnight, after which the DNA was recovered by ethanol precipitation. The precipitated DNA pellet was resuspended in 50 µl of H₂O, and PCR amplification was accomplished by the addition of 50 µl of a mixture containing 2× Vent buffer, 8 mM MgSO₄, 5 mM dNTP mix, 1 pmol of HTLV-1 oligo 2 (Fig. 1), 1 pmol of LMPCR.1 (Table 1), and 1 U of Vent DNA polymerase. These samples were placed in a thermocycler and cycled 17 times using a profile of 95°C for 1 min, 66°C for 2 min, and 72°C for 1 min, with a final extension of 10 min at 72°C. Following amplification, HTLV-1-specific PCR products were labeled by the addition of 5 µl of labeling buffer (2 mM each dNTP, 1× Vent polymerase buffer, 8 mM MgSO₄, 1 U of Vent polymerase, 2.3 pmol of an HTLV-1-specific ³²P-end-labeled oligonucleotide [Fig. 1, oligo 3]) and subjected to two rounds of amplification at 95°C for 1 min, 69°C for 2 min, and 72°C for 1 min. Each reaction product was then phenol-chloroform extracted and ethanol precipitated prior to electrophoresis on a 6% sequencing polylacrylamide gel. The reactions were visualized by autoradiography using Kodak BioMax MR film and analyzed by storage phosphor technology using a Molecular Dynamics 445 SI PhosphorImager and ImageQuant software.

Oligonucleotide sequences. A schematic of the relative positions of the oligonucleotides used for LMPCR is shown in Fig. 1. The sequences of the nested HTLV-1 primer set used for LMPCR are shown in Table 1. LMPCR.1 and LMPCR.2 are the unidirectional linkers described by Mueller and Wold (27).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

To identify the patterns of protein binding to the HTLV-1 U3 region in vivo, we used in vivo DMS footprinting and LMPCR on the HTLV-1-infected cell lines MT-2 and MT-4. In this technique, the cells were initially treated in vivo with DMS under conditions that resulted in the partial methylation of guanine residues at the N-7 position. DNA was then extracted from the cells and treated with piperidine, which specifically cleaves at methylated bases. The cleaved DNA was then subjected to a primer extension reaction using an HTLV-1-specific oligonucleotide. This resulted in double-stranded DNA terminating at guanine residues that were methylated and cleaved on the opposite strand from the primer. The double-stranded DNA was then blunt-end ligated to a double-stranded linker, thus making it amenable to amplification by nested PCR. This amplification was performed using oligonucleotides specific for the linker and the HTLV-1 U3 region. The amplified DNA was then subjected to a final round of extension using an end-labeled HTLV-1 U3-specific oligonucleotide, and the reaction products were visualized by PAGE on a DNA sequencing gel followed by autoradiography or phosphorimaging. As a control, high-molecular-weight DNA was isolated from the appropriate cell line, treated with DMS in vitro, cleaved with piperidine, and processed in parallel with the in vivo samples. If proteins occupy the DNA, then the guanine residues within the occupied region will have restricted access to DMS. This will result in a protection from methylation, reflected by diminished intensities of the corresponding bands on the autoradiograph compared to the control (in vitro-methylated) DNA. Alternatively, hypersensitivity to DNA methylation has also been noted at bases where protein is bound to DNA; this is due to increased localization of DMS created by hydrophobic pockets at the interface of globular protein domains and DNA or to alterations in the local topology of the DNA induced by protein binding. Hypersensitivity is reflected by a marked increase in the intensity of a band for in vivo-methylated DNA compared to control in vitro-methylated DNA.

Tax I activity in MT-2 and MT-4 cells. To determine the effect of Tax I on protein binding to the HTLV-1 LTR, we analyzed two infected cell lines, MT-2 and MT-4, generated by cocultivation of human cord blood with cells from patients infected by HTLV-1 (21, 26). Although both cell lines contain proviral DNA, MT-2 is a high-level producer of HTLV-1, whereas MT-4 does not produce detectable virus. The expression defect in MT-4 cells has been shown to be extensive methylation of the proviral genome (38). This defect is reversible upon treatment with 5-azacytidine, an inhibitor of methylation (38). Upon treatment with 5-azacytidine, MT-4 cells make viral RNA and protein at levels that are comparable to those produced by MT-2, suggesting that MT-4 cells contain all transcription factors necessary to transactivate the HTLV-1 LTR. There has been some question recently about the origins and behavior of MT-4 cells obtained from different sources (18). Also, there have been different reports on the amount of Tax protein made by MT-4 cells. While some investigators reported that they make very low amounts of Tax I protein (38), others reported amounts almost equivalent to those produced by MT-2 (18). To characterize the relative amounts of biologically active Tax I protein in our MT-2 and MT-4 cells, we determined their abilities to activate a Tax I-responsive reporter plasmid in a transient transfection assay. The reporter plasmid contained the SEAP gene under the control of the HTLV-1 LTR. SEAP activity was determined 72 h posttransfection and is shown in Table 2. The MT-2 cell line transactivated the HTLV-1 LTR-SEAP construct approximately 67-fold more than the MT-4 cell line. This indicated that the MT-2 cells have substantially higher levels of biologically active Tax I protein compared to the MT-2 cells.

View larger version (41K): [in this window] [in a new window]   FIG. 2.   Western blot analysis of Tax I levels in MT-2 and MT-4 cells. Whole-cell protein lysates from MT-2 and MT-4 cells were separated by SDS-PAGE and subjected to Western blot analysis using a polyclonal antibody to Tax I as described in Materials and Methods. The positions of authentic p40 Tax I as well as the Env-Tax I fusion protein made by MT-2 cells are indicated (see text for details). The band between p40 Tax I and Env-Tax I seen in both cell lines is a nonspecific band that is also observed in human T cells that do not express Tax I (not shown).

In vivo footprinting of the lower strand of the HTLV-1 TRE-1s. We first analyzed the lower (or minus) strand of the HTLV-1 U3 region in MT-2 and MT-4 cells by in vivo footprinting (Fig. 3), as it is significantly G rich. The locations of the primers (1-S plus 2a-S/3a-S or 1-S plus 2b-S/3b-S) are shown in Fig. 1. The results for the promoter-distal TRE-1 for both cell lines are shown in Fig. 3A. While MT-4 cells showed no evidence for protein occupancy within TRE-1 (no protections or hypersensitive sites), MT-2 cells showed a major protection of the central guanine residue within the CRE. In addition, MT-2 cells also had one hypersensitive and one protected G residue in the 3' GC-rich flanking sequences. We termed protections major if the reduction in cleavage at a particular residue was greater than 50%, as quantified by phosphorimaging (see below).

View larger version (58K): [in this window] [in a new window]   FIG. 3.   In vivo DMS footprinting of the lower strand of the TRE-1s. DNAs from in vivo DMS-treated MT-2 and MT-4 cells were piperidine cleaved and subjected to LMPCR as described in Materials and Methods. The reaction products were separated on polyacrylamide gels and exposed to film. Autoradiograms corresponding to the three TRE-1s are shown. (A) Promoter-distal TRE-1. A band is missing in the 5' GC-rich flank most likely because the proviral DNA contains a base other than the predicted guanine residue at that position. (B) Promoter-central TRE-1. (C) Promoter-proximal TRE-1. Arrowheads pointing toward bands indicate hypersensitive residues; arrowheads pointing away from bands indicate protected residues. Larger arrowheads indicate major protections or hypersensitive sites.

View larger version (21K): [in this window] [in a new window]   FIG. 4.   PhosphorImager quantitation of the MT-2 lower-strand promoter-central TRE-1. The gel in Fig. 3B was exposed to a storage phosphor screen. The image was scanned, and the graph was generated by using ImageQuant 5.0 software. The protected and hypersensitive residues shown in Fig. 3B are indicated.

In vivo footprinting of the upper strand of the HTLV-1 TRE-1s. Analysis of the upper (or plus) strand of the HTLV-1 U3 region is shown in Fig. 5. Due to the distance of the nested primers (2a-AS and 3a-AS) from the TRE-1s (Fig. 1), we were unable to resolve fragments for the upper strand of the promoter-distal TRE-1. Two other sets of primers failed to resolve this element. The footprint of the promoter-central TRE-1 is shown in Fig. 5A. This repeat has three G residues within the CRE on the upper strand. In MT-2 cells, the first two guanine residues were protected (major protection of the central G), while MT-4 cells did not shown any protections or hypersensitivities within this region. Analysis of the promoter-proximal TRE-1 is shown in Fig. 5B. In MT-2 cells, both guanine residues within the CRE were strongly protected. Additionally, all three G residues within the 5' GC-rich flank were strongly protected. Again, the MT-4 cell line did not show any appreciable DNA-protein interactions within this element. The results of the upper-strand analysis were consistent with results of the lower-strand analysis in that there were protections on the CRE in the central and proximal TRE-1s, but only in the MT-2 cells.

View larger version (37K): [in this window] [in a new window]   FIG. 5.   In vivo DMS footprinting of the upper strand of the promoter-central (A) and promoter-proximal (B) TRE-1s. In vivo DMS LMPCR was performed on MT-2 and MT-4 cells as described for Fig. 3.

View larger version (20K): [in this window] [in a new window]   FIG. 6.   Summary of TRE-1 footprints in the MT-2 cell line. The footprint data presented in Fig. 3 and 5 are summarized on the individual TRE-1 sequences. The CRE is in bold. "H" indicates hypersensitivity, and "P" indicates protection; bolded P's and H's reflect strong interactions. The upper strand of the promoter-distal TRE-1 is italicized to indicate that it was not analyzed.

In vivo footprinting on the HTLV-1 TRE-2 sequences. We also analyzed the TRE-2 region between the promoter-central and promoter-proximal TRE-1s. This region contains potential binding sites for a number of transcription factors, including Sp1, NFB, the Ets family proteins PEA-3 and PU.1, HNF-3, and AP-2 (see Fig. 8). The NFB and Sp1 sites are contained in the same DNA sequences on opposite strands, and the same is true for the PU.1 and PEA-3 elements. Analysis of the lower strand of the TRE-2 is shown in Fig. 7A. In MT-4 cells, hypersensitivity of two of the five G residues within the predicted Sp1/NFB binding site was observed. Also, three consecutive G residues in the putative AP-2 binding site were mildly protected. In contrast, the MT-2 cells demonstrated a stronger and more extensive binding pattern compared to the MT-4 cell line. In MT-2 cells, all five G residues in the predicted Sp1/NFB binding site were protected, and a number of residues were strongly protected. In addition, all G residues in the PEA-3 and AP-2 elements were either hypersensitive or protected, and the protections over the AP-2 site were particularly strong.

View larger version (86K): [in this window] [in a new window]   FIG. 7.   In vivo DMS footprinting of TRE-2 in the MT-2 and MT-4 cell lines. In vivo DMS LMPCR was performed on MT-2 and MT-4 cells. (A) Lower strand of TRE-2 in MT-2 and MT-4 cells. The predicted binding sites for different transcription factors for this strand are also indicated. Some transcription factor binding sites overlap other binding sites on the opposite strand (Fig. 8). (B) Upper strand of TRE-2 in MT-2 and MT-4 cells.

View larger version (21K): [in this window] [in a new window]   FIG. 8.   Summary of TRE-2 footprints in the MT-2 and MT-4 cell lines. The footprint data presented in Fig. 6 are summarized on the TRE-2 sequence. Predicted transcription factor binding sites are shown in bold. Conventions are the same as for Fig. 6.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, we carried out in vivo footprinting on the HTLV-1 LTR in infected T-cell lines. These experiments provided insight into the binding of factors to different motifs in the HTLV-1 LTR in the context of integrated proviruses. The results provide new information that complements the previous extensive in vitro experiments on binding of cellular factors to HTLV-1 DNA. In particular, it was possible to test if the protein-DNA interactions predicted from the in vitro binding experiments were observed in vivo. By comparing MT-2 and MT-4 cells, it was possible to study factor binding to the HTLV-1 LTR in cells with or without functional Tax I protein.

The most interesting result that emerged from this study was the fact that there was no evidence for factor binding to the TRE-1s (21-bp repeats) in MT-4 cells, while there was binding in MT-2 cells. Thus, in vivo, Tax I appears to be required for binding of cellular factors to the TRE-1s. This was particularly noteworthy, since in vitro experiments have indicated that factors (bZIP proteins of the CREB/ATF superfamily) can bind to TRE-1 DNA (the central CRE in particular) in the presence or absence of Tax I (17, 30, 41, 44, 50). The in vivo Tax I-dependent binding might reflect a higher affinity for the binding of these factors to the central CRE in the presence of Tax I. Indeed, increased affinity of CREB for TRE-1 DNA in the presence of Tax I has been reported (5, 47, 48, 51).

As indicated above, each TRE-1 contains a central CRE, and Tax I protein can form an in vitro complex with CREB and TRE-1 DNA (5, 51). Methylation interference footprinting of the in vitro CREB-TRE-1 or Tax I-CREB-TRE-1 complexes has shown that CREB contacts all three GC base pairs within the central CRE motif (TGACG) (32). In MT-2 cells, the in vivo footprinting showed strong protection of all G's in the TRE-1 CREs, consistent with binding of CREB in vivo. However, it is possible that other members of the CREB/ATF family of transcription factors (either as homodimers or heterodimers) could also bind and protect the G's in the central CREs. It has also been shown that Tax I extends the in vitro footprint in a CREB-TRE-1 complex by contacting the minor groove of the GC-rich flanking sequences (22). The in vivo footprinting of the TRE-1s in MT-2 cells showed additional protected G residues in the sequences flanking the central CREs, particularly for the promoter-proximal TRE-1. Thus, the in vivo footprints of the TRE-1s in MT-2 cells were generally consistent with patterns predicted from the in vitro footprinting experiments.

It was noteworthy that while the central CREs were strongly protected in the in vivo footprints of all three TRE-1s in MT-2 cells, the extents of protection in the GC-rich flanking sequences differed. The most extensive footprints in the flanking sequences were observed for the promoter-proximal TRE-1 (all G residues protected or hypersensitive), the central TRE-1 showed intermediate protein interactions (some G residues not protected), and the distal TRE-1 showed minimal protections in the flanking sequences (although only the lower strand could be analyzed) (Fig. 6). One possible explanation for this could be that the different TRE-1s bind CREB or Tax I in vivo but with different affinities, resulting in differential spreading of the footprints to the flanking sequences. The different affinities could result from differences in nucleotide sequences or from relative proximity to the basal promoter elements. Alternatively, different protein complexes might bind the different TRE-1s due to sequence differences. It has been observed that the different TRE-1s can form qualitatively different protein-DNA complexes in gel shift assays (43, 46).

We also analyzed in vivo footprints over the TRE-2 sequences. In contrast to the TRE-1s, where footprints were observed only in MT-2 cells, footprints were observed in both MT-2 and MT-4 cells, although the patterns were quite different. In MT-4 cells, the main footprint was characterized by one strong and one weak hypersensitive site in the NFB/Sp1 region (as well as weak protection over the AP-2 site), while most other G residues in the TRE-2 were not protected. In contrast, in MT-2 cells, there were extensive footprints over the TRE-2 sequences; indeed, all G residues in the identified motifs in this region were protected (or hypersensitive), and only three G residues in the entire region were not protected. Thus, while there may be limited factor binding to the TRE-2 region in the absence of Tax I, the transactivator protein leads to extensive factor binding to this region. Even for the NFB/Sp1 region, where there was evidence for factor binding in both MT-2 and MT-4 cells, it was clear that different proteins were bound. In MT-4 cells, the footprint consisted of hypersensitive sites in two of five G residues; in MT-2 cells, four of the five G residues were protected, including the G residue that was strongly hypersensitive in MT-4 cells. The factors bound to the TRE-2 sequences in MT-2 cells remain to be elucidated. It will be important to couple these in vivo studies with additional in vitro experiments (e.g., antibody supershift experiments).

In vivo footprinting indicated that the strongest protein-DNA interactions were for the promoter-proximal TRE-1 and TRE-2, while weaker or more limited interactions were observed for the distal and central TRE-1s. This might suggest that the promoter-proximal TRE-1 and TRE-2 are the critical elements for transcriptional activity of the HTLV-1 LTR in MT-2 cells. Indeed, it has been shown that together these two elements are sufficient for Tax I responsiveness of the HTLV-1 LTR (24, 28).

These studies were carried out on cell lines that contain multiple copies of HTLV-1 provirus (some defective) integrated into the genome (6, 20). The in vivo footprinting described here is an averaged analysis of all of the HTLV-1 proviruses in a given cell line, regardless of whether they were transcriptionally active or inactive. If inactive proviruses in MT-2 cells contained fewer or no bound factors, this would have reduced the apparent strength of protections or hypersensitivities in the active proviruses in the footprints. Conversely, particularly strong protections or hypersensitive sites might have originated from both active and inactive proviruses.

It should be noted that due to locations of the LMPCR primers, analysis of the lower strand involved amplification from both the upstream and downstream LTRs. In contrast, the primers used for analysis of the upper strand would specifically amplify fragments for the upstream LTR. (It was not possible to design primers that would exclusively analyze the lower strand in the upstream LTR, since this would require knowledge of the flanking host cell sequences for each provirus.) Transcription for the virus initiates in the upstream LTR and terminates in the downstream LTR, and it has been shown that the downstream LTR is less transcriptionally active (14). It seems quite possible that transcription factors might be loaded on the upstream LTR but not on the downstream LTR. It was interesting that protections on the upper strand were generally stronger than those on the lower strand (e.g., compare Fig. 3 and 5). This would be consistent with differential loading of factors on the upstream and downstream LTRs.

In summary, this report represents the first detailed in vivo footprint analysis of the HTLV-1 LTR. The results indicated Tax I-dependent loading of factors onto both the TRE-1s and TRE-2. Within the TRE-1s, there was evidence for factor binding at the central CREs, as well as binding to flanking GC-rich sequences. Moreover, the strongest footprints were observed for the promoter-proximal TRE-1 and TRE-2. These results confirm and extend prior in vitro footprinting experiments on the HTLV-1 LTR.