Naturally Occurring Human Immunodeficiency Virus Type 1 Long Terminal Repeats Have a Frequently Observed Duplication That Binds RBF-2 and Represses Transcription

Abstract

Approximately 38% of human immunodeficiency virus type 1 (HIV-1)-infected patients within the Vancouver Lymphadenopathy-AIDS Study have proviruses bearing partial 15- to 34-nucleotide duplications upstream of the NF-κB binding sites within the 5′ long terminal repeat (LTR). This most frequent naturally occurring length polymorphism (MFNLP) of the HIV-1 5′ LTR encompasses potential binding sites for several candidate transcription factors, including TCF-1α/hLEF, c-Ets, AP-4, and Ras-responsive binding factor 2 (RBF-2) (M. C. Estable et al., J. Virol. 70:4053–4062, 1996). RBF-2 and an apparently related factor, RBF-1, bind to at least four cis elements within the LTR which are required for full transcriptional responsiveness to protein-tyrosine kinases and v-Ras (B. Bell and I. Sadowski, Oncogene 13:2687–2697, 1996). Here we demonstrate that representative MFNLPs from two patients specifically bind RBF-2. In both cases, deletion of the MFNLP caused elevated LTR-directed transcription in cells expressing RBF-2 but not in cells with undetectable RBF-2. RBF-1, but not RBF-2, appears to contain the Ets transcription factor family member GABPα/GABPβ1. Taken together with the fact that every MFNLP from a comparative study of over 500 LTR sequences from 42 patients contains a predicted binding site for RBF-2, our data suggest that the MFNLP is selected in vivo because it provides a duplicated RBF-2 cis element, which may limit transcription in monocytes and activated T cells.

The cellular trans-acting factors which are thought to regulate transcription from the human immunodeficiency virus type 1 (HIV-1) long terminal repeat (LTR) were determined from some of the first molecular isolates (57, 63, 64) which have been extensively characterized and consequently are considered prototypical (12, 25–27, 36). The prototype LTR includes LTR/Nef-coding sequences (−454 to −121) and LTR/noncoding sequences (−120 to +80), which appear to be evolving independently in vivo (14, 22, 49) (Fig. 1). The LTR/Nef-coding region of the prototypical HIV-1 LTR consists of a negative regulatory region (−340 to −185) (7, 57) and an upstream regulatory element (URE) (−157 to −122) (54). The negative regulatory region contains binding sites for the nuclear factor of activated T cells (NF-AT) (47, 61) and upstream stimulatory factor (17), while the URE includes binding sites for lymphoid enhancer-binding factor (hLEF; also referred to as TCF-1α) (62, 66, 70), Ets-1 (34, 62), and Ras-responsive binding factors 1 and 2 (RBF-1 and -2), which bind to Ras-responsive binding elements (RBE) IV (−151 to −142) and RBE III (−131 to −122) (6, 22) (Fig. 1). However, in vivo the only highly conserved transcription factor DNA-binding motifs in the LTR/Nef-coding region appear to be RBE III and RBE IV/Ets (6, 22, 52).

FIG. 1.

Schematic representation of frequently detected naturally occurring HIV-1 LTR proviral length polymorphisms. The approximate locations of length polymorphisms (deletions and insertions) observed in HIV-1 LTRs from AIDS patients, including the most frequent naturally occurring length polymorphism (MFNLP) at the boundary (−121/−120) between the Nef-coding region (Nef LTR DNA) and noncoding region (LTR DNA) of the LTR, are indicated. Positions of RBE I, RBE II, RBE III, and RBE IV [see the introduction]) are indicated with respect to the typical modulatory, enhancer, basal promoter, and TAR regions. Binding sites for Ets, hLEF/TCF-1α, NF-κB, SP-1, RBF-1, and RBF-2 are indicated.

The cis-acting elements of the LTR/noncoding region include an enhancer (−104 to −81), a basal promoter (−80 to +40), and a Tat-responsive region (TAR) (+1 to +59) that is transcribed into an RNA stem-loop structure involved in strong activation of transcription from the HIV-1 LTR (27, 30). Transcription factors binding the enhancer region (−104 to −81) include members of the Rel/kappa B (NF-κB) family (2, 28, 51, 53, 59) and Ets family members (23, 32, 60), including RBF-1 (6), that bind to Ets sites embedded in the 3′-half sites of the NF-κB motifs. For this reason, we have termed the −80 to −104 region RBE II, to indicate that it binds RBF-1 as well as other Ets family members (6) (Fig. 1). Factors which bind to the basal promoter elements include SP-1 family members (19, 20, 37, 48), TATA-binding protein (26), E-box factors (56), and RBF-2, which binds to RBE I (−26 to −5) (6, 22) (Fig. 1). Additional host factors appear to bind the TAR region (27, 30). Within the LTR/noncoding region, although the enhancer region is highly conserved in vivo (14, 22, 41, 49, 50, 52), naturally occurring variants with point mutations that are predicted to impair NF-κB, Ets, and RBF-1 binding have been detected (22). Indeed, in at least some assays, mutations to one or deletion of both NF-κB sites does not abrogate viral replication (8, 21, 44, 58, 67). Moreover, a pathogenic HIV-1 isolate that completely lacks the prototypical enhancer sequences has recently been described (73). In vivo, mutations to the basal promoter which are predicted to impair binding of SP-1, TATA-binding protein, and E-box-binding proteins, and mutations to TAR and non-TAR DNA that impair Tat transactivation, have also been described (22).

Natural HIV-1 LTRs, including those from AIDS patients, also contain length polymorphisms (Fig. 1). These include the insertions and deletions which are summarized in Fig. 1 (1, 22, 29, 41, 50, 73). In particular, a length polymorphism immediately 5′ of the enhancer region in which sequences overlapping the hLEF/TCF-1α site are duplicated has been detected by several groups (1, 22, 29, 39, 41, 42, 50, 72, 73). Several recent publications refer to these insertions as partial TCF-1α sites (50, 72, 73). Golub and coworkers noted that one such duplication increased viral replication and transcription in the context of an additional mutation between the NF-κB motifs (29) and pointed out that sequences within the duplication corresponded to unduplicated prototype LTR sequences (−139 to −119) previously noted to increase phorbol ester responsiveness (38). This same region, unduplicated (−157 to −122), has been described by Nakanishi and coworkers as the URE, acting positively in MOLT-4 or U937 cells but as a negative regulatory element in MT-4 or Jurkat cells (54). These authors also demonstrated three specific complexes (URE-binding factors) formed on this region with nuclear extracts from HeLa cells (54). Koken and coworkers have described this same duplication as the CTG motif and were the first to note that most of the duplications in this region contained a 5′-ACTGCTGA-3′ sequence that is also present in HIV ANT-70 and simian immunodeficiency virus (41). These authors demonstrated that this motif enhanced the positive effect of the NF-κB sites on LTR-directed transcription in HeLa or Jurkat cells and suggested that it binds a 68-kDa nuclear protein (42). However, the duplicated CTG motif was observed to have a slight negative effect on transcription and replication in vivo (41, 42).

Recently, we have shown that this polymorphism is the most frequent naturally occurring length polymorphism (MFNLP) of the HIV-1 LTR, present in 38% of sampled patients, and that its occurrence does not correlate with the clinical laboratory parameters of CD4 count, stage, duration of infection, and progression rate (22). Whereas MFNLPs contain partial or full TCF-1α duplications, we have also noted that they invariably duplicate RBE III, representing a binding site for RBF-2, and that some MFNLPs additionally contain potential Ets GGA core binding site sequences. Because there is a correlation between MFNLP occurrence and mutations to the RBE sites, we have proposed a compensatory role for this frequent polymorphism (6, 22).

Resolving the function of this duplication and identification of the transcription factor(s) that it binds are important for our understanding of HIV-1 transcription in vivo, particularly in light of a recently described pathogenic HIV-1 isolate which lacks an enhancer but which contains the MFNLP duplication (73). We demonstrate here that MFNLPs bind a specific complex from Jurkat nuclear extracts which is indistinguishable from RBF-2. Additionally, we show that hLEF and Ets proteins do not interact with all MFNLPs. Using isogenic constructs, we demonstrate that the MFNLP has a repressive effect on HIV-1 LTR transcription in Jurkat cells but not in cells lacking RBF-2. Finally, we show that RBF-1, a factor which binds to sites which may be compensated for by the MFNLP, contains the Ets family transcription factor GABP. Taken together, our data argue for in vivo MFNLP selection based on RBF-2 binding and not hLEF/TCF-1α or Ets family proteins.

MATERIALS AND METHODS

EMSA.

Oligonucleotides were synthesized on an Applied Biosystems 391 DNA synthesizer. After cleavage from the column and deprotection, the oligonucleotides were dried in a SpeedVac, purified on Sep-Pak columns, and stored in H₂O at 100 pmol/μl; 100 pmol of annealed oligonucleotide, bearing 5′ overhangs, was labeled by end filling with Klenow enzyme and [α-³²P]dATP (NEN). Labeled oligonucleotides were precipitated with ethanol, washed three times with 70% ethanol, and dried in a SpeedVac. Jurkat nuclear extracts were prepared as previously described (6). Antibodies to GABPα, -β1, and -β2 were a kind gift from Steve McKnight and Fabienne de la Brousse (Tularik Inc.). Electrophoretic mobility shift assay (EMSA) binding reactions consisted of 100 pmol of ³²P-labeled double-stranded target oligonucleotide, 10 mM HEPES (pH 7.9), 5 mM MgCl₂, 8% glycerol, 100 mM KCl, 6 μg of poly(dI-dC), 4 μg of bovine serum albumin, and the competitor double-stranded DNA (45). The radiolabeled probe was always added last, immediately following addition of nuclear extract. Samples were kept on ice until the radiolabeled double-stranded oligonucleotides were added and were then incubated at room temperature for 20 min and resolved on 4.5% acrylamide–bisacrylamide (29:1)–0.5× Tris-borate-EDTA–1% glycerol 0.8-mm-thick gels.

Synthetic double-stranded oligonucleotides used in EMSA experiments are shown in Fig. 2. RBE IV, RBE IIIL, RBE IIIL Mut1, RBE IIIS, RBE IIIS Mut1, RBE IL, and RBE IS are equivalent to oligonucleotides A (wt), P3, P3M1, C, mC2, PT, and F, respectively, as described previously (6).

FIG. 2.

MFNLP sequences and synthetic oligonucleotides used for EMSA. MFNLP-A and MFNLP-B are double-stranded oligonucleotides representing the 31-nucleotide duplication from pMCE 9.104 and the 24-nucleotide duplication from pMCE 69.1, respectively. RBEs, double-stranded oligonucleotides representing the RBE III element (RBE IIIS, short version; RBE IIIL, long version), the RBE I element (RBE IS, short version; RBE IL, long version), and the RBE IV element of wild-type HIV-1 LTR sequences. Nucleotide substitutions within mutant (Mut) oligonucleotides are shaded. Other, double-stranded oligonucleotides representing a binding site for AP-4. RBEs are indicated on wild-type oligonucleotides by dashed boxes, as are potential binding sites for AP-4, hLEF, and Ets.

DNase I footprinting.

pMCE clones (22) were digested with HindIII to linearize and labeled with Klenow enzyme and [α-³²P]dATP to a specific activity of 2 to 10,000 cpm/fmol. The labeled DNA was then digested with XbaI, and the α-³²P-labeled 420-bp LTR fragments were purified by agarose gel electrophoresis and chromatography on NACS columns (Bethesda Research Laboratories). Recombinant c-Ets-1 DNA-binding domain (residues 300 to 440; Ets-1Δ301) was a kind gift from Logan Donaldson and Lawrence MacIntosh. Purified hLEF was a kind gift from Marianne Waterman (University of California, San Francisco).

Approximately 5 to 10 fmol of radiolabeled LTR fragments was preincubated on ice for 30 min in 40 μl containing 10 mM HEPES (pH 7.9), 5 mM MgCl₂, 8% glycerol, 100 mM KCl, 5 μg of preannealed poly(dI-dC) (Pharmacia), and the relevant DNA-binding protein. Five microliters of DNase I (Promega), freshly diluted (to the appropriate concentrations) in 150 mM NaCl–1 mM CaCl₂–50% glycerol, was added to the binding reactions; the tubes were vortexed and incubated at room temperature for 2 min. Digestions were terminated by the addition of 60 μl of 200 mM NaCl–20 mM EDTA–1% sodium dodecyl sulfate–250 μg of yeast tRNA per ml–150 μg of proteinase K per ml (Boehringer Mannheim). The samples were mixed and incubated at 42°C for 60 min. Template DNA was extracted with phenol-chloroform (50:50, vol/vol) and precipitated with ethanol. Maxam-and-Gilbert (M&G) A+G ladders were generated by standard protocols. Template DNA samples were suspended at 2,000 to 10,000 cpm/μl, and equal counts were resolved by electrophoresis on denaturing 6% polyacrylamide gels. Gels were dried and exposed to X-Omat (Kodak) film with an enhancing screen (Dupont).

Site-directed mutagenesis.

Oligonucleotide-directed mutants were generated by primer extension on uridine-substituted template DNA, using the Kunkel modification (43). Synthetic oligonucleotides spanning the pMCE 69.1 (5′-GAATACTACAAGAACTGAACTCATCGAGCTTTCTACAAG-3′) and pMCE 9.104 (5′GAATTCTACAAGAACTGATGACACTGAGCTATCTACAAGGGAC- 3′) MFNLP-flanking sequences were used to create pMCE Δ 69.1 and pMCE Δ 9.104, respectively. Oligonucleotides were phosphorylated with T4 polynucleotide kinase and stored at −20°C. Uridine-substituted single-stranded pMCE 69.1 and pMCE 9.104 were prepared by passage through Escherichia coli CJ236, using M13K07 helper phage (Pharmacia). For each reaction, 3 μg of single-stranded DNA was annealed to the kinase-treated oligonucleotide, and the final synthesis-mutagenesis reaction product was transformed into E. coli DH5α cells. Colonies were picked, and deletion mutant LTR DNAs were identified by sequencing (Applied Biosystems 373 DNA sequencer) as outlined elsewhere (22).

Transfection and CAT assays.

Cells were transiently transfected by using a DEAE-dextran technique which has been described elsewhere along with our protocol for chloramphenicol acetyltransferase (CAT) assays (22).

RESULTS

Two different MFNLPs form a specific complex with the same nuclear factor from Jurkat cells.

To examine the cellular factors that may bind to representative MFNLPs, we selected clones pMCE 9.104 and pMCE 69.1 from our collection of group M, subtype B (23) HIV-1 LTRs isolated from Vancouver Lymphadenopathy-AIDS study (VLAS) samples (22) for further analysis. These two clones contain MFNLP duplications of 31 and 24 nucleotides, respectively (Fig. 1), and were chosen because they have potential AP-4, hLEF, Ets, and RBE III sites, thus representing sequences for all of the putative transcription factor binding sites that we have been able to identify within MFNLPs (22). Oligonucleotides representing the MFNLPs from pMCE 9.104, designated MFNLP-A, and pMCE 69.1, designated MFNLP-B (Fig. 2), were used as probes in EMSA with Jurkat cell nuclear extracts to determine whether these duplications formed specific complexes with proteins. We found that MFNLP-A formed several complexes with Jurkat nuclear proteins (Fig. 3). The band labeled S was found to represent a specific protein–MFNLP-A complex because this band could be eliminated by inclusion of excess unlabeled MFNLP-A oligonucleotide in the binding reaction (Fig. 3; compare lane 2 with lanes 3 to 6), whereas the nonspecific band was largely unaffected by excess competitor oligonucleotide (lanes 3 to 6). This MFNLP-A complex requires the TGA motifs within MFNLP-A, because mutations to either the first (lanes 7 to 9) or both (lanes 13 to 15) TGA motifs of the competitor oligonucleotide (also see Fig. 2) reduced competition. The affinity of this complex for the first TGA motif appears to be significantly greater than for the second, because mutation to the second TGA alone does not significantly impair competition (lanes 10 to 12).

FIG. 3.

MFNLP-A binds a specific factor from Jurkat nuclear extracts, as determined by EMSA with radiolabeled synthetic MFNLP-A (Fig. 1). Reactions contained no extract (lane 1) or 2.5 μg of Jurkat nuclear extract (lanes 2 to 26) and 1 pmol of radiolabeled oligonucleotide. Unlabeled competitor (Comp) oligonucleotides were added to the binding reactions, as indicated above each lane, at 50-fold (lanes 4, 7, 10, and 13), 100-fold (lanes 5, 8, 11, and 14), or 200-fold (lanes 3, 6, 9, 12, and 15 to 26) molar excess. S, specific complex; NS, nonspecific complex.

The oligonucleotide derived from pMCE 69.1 (22), designated MFNLP-B (Fig. 2), was also found to form a specific complex with proteins from Jurkat nuclear extracts, since the band labeled S in Fig. 4 could be competed with a 200-fold molar excess of unlabeled MFNLP-B, whereas the nonspecific band was unaffected by competitor oligonucleotide (Fig. 4; compare lanes 2 and 3). This specific MFNLP-B complex was found to require the second TGA motif within MFNLP-B, because mutations to these nucleotides prevented competition (lane 6), whereas mutations to the first TGA (lane 4) or a non-TGA motif (lane 5) did not (also see Fig. 2).

FIG. 4.

MFNLP-B binds a specific factor from Jurkat nuclear extracts, as determined by EMSA with radiolabeled synthetic MFNLP-B (Fig. 1). Reactions contained no extract (lane 1) or 2.5 μg of Jurkat nuclear extract (lanes 2 to 10) and 1 pmol of radiolabeled oligonucleotide. Unlabeled competitor (Comp) oligonucleotides as indicated above the lanes were added to the binding reactions, at 200-fold molar excess (lanes 3 to 10). S, specific complex; NS, nonspecific complex.

We also found that the specific complex formed with the oligonucleotide representing MFNLP-A could be competed with excess heterologous oligonucleotide MFNLP-B (Fig. 3, lane 16), suggesting that these duplications from two different patients specifically bind the same nuclear factor. Consistent with this notion, we found that mutation of the second TGA motif of MFNLP-B inhibited its ability to compete for binding the MFNLP-A-specific complex (lane 19). In contrast, as we observed with EMSA experiments using an MFNLP-B probe, mutations to the first TGA (lane 17) or a non-TGA motif (lane 18) of the MFNLP-B competitor had little effect on competition for specific complex formation with MFNLP-A. Conversely, we also found that the specific complex formed with the oligonucleotide representing MFNLP-B could compete with unlabeled excess heterologous oligonucleotide MFNLP-A (Fig. 4, lane 7). These results demonstrate that two different MFNLPs specifically bind an identical nuclear factor which requires the TGA motifs for DNA binding.

The TGA-specific complexes formed with MFNLP-A and -B can be competed by oligonucleotides recognizing RBF-2.

We found that the TGA-specific complex formed with MFNLP-A could be eliminated by competitor oligonucleotides containing RBF-2 binding sites. Specifically, oligonucleotides representing RBE I (HIV-1 LTR nucleotides −26 to −5 [Fig. 1]) (Fig. 3, lane 26) and RBE III (HIV-1 LTR nucleotides −131 to −122 [Fig. 1]) (lanes 22 and 23) could compete with MFNLP-A for binding the TGA-specific complex. In contrast, neither an RBE III oligonucleotide containing a mutation of its TGA motif (lane 24) nor an oligonucleotide containing only a partial RBE I site (lane 25) could compete for binding the MFNLP-A complex. Also, an oligonucleotide containing a binding site for RBF-1/Ets (RBE IV [lane 21]) was unable to compete for MFNLP-A-specific complex formation. Similarly, although most MFNLPs contain sequences resembling an AP-4 site, we found that a competitor oligonucleotide containing a strong consensus AP-4 site (Fig. 2) was unable to compete for the MFNLP-A complex (Fig. 3, lane 20). Consistent with this result, we found that antibodies against AP-4 protein do not interfere with formation of this complex (not shown). Similarly, the TGA-specific complex formed with MFNLP-B can be competed with an RBE III site (Fig. 4, lane 9) but not an RBE IV site (lane 8) or a shortened RBE I site (lane 10). These results demonstrate that the protein complex which binds the MFNLP from two different patients appears to have binding specificity similar to that of the previously described factor RBF-2 but is distinct from AP-4 or RBF-1/Ets.

MFNLP-A and -B compete for binding of RBF-2 to RBE III.

We have previously defined RBF-2 as a nuclear factor which binds at least two sites within the HIV-1 LTR termed RBE III and RBE I (6). Therefore, RBF-2 can be observed as a specific complex which forms with labeled RBE III (Fig. 5, lane 1) that can be competed with unlabeled RBE III (lane 2) or unlabeled RBE I (reference 6 and not shown). The specific RBF-2 complex cannot be competed with oligonucleotides which bind RBF-1/Ets (RBE IV [lane 4]) or a strong consensus AP-4 binding site (lane 5). RBF-2 is immunologically distinct from hLEF, and its DNA-binding component is approximately 100 kDa, as determined by Southwestern blotting and UV cross-linking (6). In addition, we have previously observed that every MFNLP within the VLAS contains a predicted duplication of the RBF-2 binding site represented by RBE III (22). Also, 86% of the MFNLPs within the VLAS correlate with the co-occurrence of mutations to RBE sites (22). Therefore, to confirm that the MFNLPs bind RBF-2, we examined whether MFNLP-A and MFNLP-B oligonucleotides could compete for binding of RBF-2 to a labeled RBE III oligonucleotide in EMSA. We found that inclusion of excess MFNLP-A (Fig. 5, lane 6) or MFNLP-B (lane 10) in the binding reaction inhibited specific complex formation with RBE III as efficiently as did an unlabeled RBE III competitor (lane 2). Mutations to the first or both TGA motifs within MFNLP-A prevent competition for RBF-2 (lanes 7 and 9). Similarly, mutation of the second TGA motif in MFNLP-B prevents its ability to compete for RBF-2 binding (lane 12). Therefore, in combination with the results shown in Fig. 3 and 4, these experiments demonstrate that MFNLPs from two different patients bind a factor which is likely identical to RBF-2, a result which is consistent with the fact that these polymorphisms duplicate sequences which overlap RBE III.

FIG. 5.

The specific complex formed by MFNLPs and Jurkat nuclear extracts is indistinguishable from RBF-2, as determined by EMSA with radiolabeled synthetic RBE and Jurkat nuclear extracts. Unlabeled competitor (Comp) oligonucleotides as indicated above lanes 2 to 13 were added at 200-fold molar excess. The position of the RBF-2 complex is indicated. NS, nonspecific complex.

Most MFNLPs do not bind Ets family members or hLEF.

Because some MFNLPs contain a GGA(A/T) sequence, representing a potential Ets-like core binding element (24, 32, 55, 60), we wished to determine whether these duplications caused the introduction of additional binding sites for an Ets protein family member within the LTR. To examine this possibility, we used recombinant protein representing the DNA-binding domain of c-Ets-1 (Ets-1Δ301) for in vitro footprinting experiments with naturally occurring HIV-1 LTRs (22). For these experiments we used a collection of LTR templates which contain different MFNLPs (pMCE 9.104 [MFNLP-A], pMCE 16.1, pMCE 25.5, pMCE 59.1, and pMCE 69.1 [MFNLP-B]) (22), as well as clones pMCE 4.81 and pMCE 36.1, which lack MFNLPs (22). As expected, for most of the LTR templates we observed binding of recombinant Ets protein to the enhancer region, which contains Ets sites embedded within the NF-κB elements (termed RBE II), and to the upstream RBF-1/Ets site RBE IV. However, pMCE 9.104 and 59.1 have mutations within RBE IV and II, respectively, and therefore do not bind recombinant Ets at these locations (Fig. 6A, 9.104 and 59.1). The polymorphism to the RBE II site of pMCE 59.1 also abolished binding of the p50 subunit of NF-κB to the 5′ NF-κB motif (not shown). Among the five LTRs containing MFNLPs, only pMCE 25.5 and pMCE 69.1 (22) were found to bind Ets protein within the duplication (Fig. 6A). This result demonstrates that the MFNLPs are likely not selected in vivo for their ability to bind Ets family members. It is also interesting that several of the naturally occurring LTRs, including pMCE clones 4.81, 9.104, 16.1, and 25.5, have additional, previously unreported Ets binding sites (Fig. 6A), some of which are a direct consequence of a deletion in this region indicated in Fig. 1 (22).

FIG. 6.

Recombinant Ets and hLEF do not bind most MFNLPs. (A) DNase I footprinting analysis of the HIV-1 LTRs from the pMCE clones (22) as indicated above the lanes and recombinant Ets-1 DNA-binding-domain protein. The undigested labeled probe (2 to 50 fmol; lanes 1), G+A M&G sequencing reaction (lanes 2), and probe digested with 1 U of DNase I in the absence (lanes 3) or presence (lanes 4) of 4 μg (24 pmol) of recombinant Ets-1 Δ-301 protein were analyzed on 6% denaturing gels. The positions of RBE I, RBE II, RBE III, and RBE IV are indicated at the left; Ets binding sites (EBS) are indicated with grey bars on the right. (B) DNase I footprinting analysis of the HIV-1 LTRs from the pMCE clones (22) as indicated above the lanes and recombinant hLEF. The G+A M&G sequencing reaction (lanes 1) and probe digested with 0.1 U (lanes 2) or 1 U (lanes 3) of DNase I or in the presence of 0.2 μg of recombinant hLEF (lane 4) were analyzed on 6% denaturing gels. The sequences of potential hLEF binding sites are indicated on the right; those which are not protected by hLEF are noted with an “X.”

Because several recent publications have referred to the MFNLP as a partial TCF-1α/hLEF duplication (50, 72, 73), we examined whether these duplications were capable of binding recombinant hLEF in vitro. Surprisingly, we could not detect binding of recombinant hLEF to the MFNLPs on LTR clone 59.1, 69.1, or 9.104 (Fig. 6B). We did observe binding of hLEF to its predicted element upstream of the MFNLP on LTR clone 59.1 but not on LTR clone 36.1, 69.1, or 9.104 (Fig. 6B). We believe that these results reflect the fact that although the prototypical HIV-1 LTR represented by the HXB2 clone contains a strong upstream consensus binding site for a TCF-1α/hLEF site (CAAAG) (18, 46), many of the LTR sequences that we obtained from HIV-1-infected individuals and AIDS patients (as well as those in the Los Alamos database [52]) have the sequence CAAGA (22). Since high-mobility-group proteins, such as hLEF, bind to DNA in the minor groove and the AAA bases of the high-affinity hLEF site form part of this groove (46), the A-to-G difference could explain the poor binding of hLEF that we detected. Therefore, these results suggest that most MFNLPs do not bind hLEF. In fact, our analysis of LTR sequences from HIV-1-infected patients indicates that only 31% have prototypical hLEF sites (22), and only rare MFNLP duplications would be predicted to provide a strong binding site for hLEF/TCF-1α (22), based on its DNA-binding specificity.

The MFNLP inhibits HIV-1 LTR-directed transcription in cells expressing RBF-2.

To determine the contribution of the MFNLP to HIV-1 LTR-directed transcription, we created variants of the pMCE 9.104 and pMCE 69.1 LTR clones (22) in which the MFNLPs were precisely removed by site-directed mutagenesis. We found that in the Jurkat T-cell line, removal of MFNLP-A from pMCE 9.104 caused a significant 75% increase in LTR transcription, while removal of MFNLP-B from pMCE 69.1 caused a 25% increase (Fig. 7A). We also found that the observed apparent repressive effect of the MFNLP on LTR-directed transcription was greatly accentuated in Jurkat cells that were cotransfected with a plasmid expressing HIV-1 Tat protein (Fig. 7B). Thus, the parent pMCE 69.1 plasmid produced approximately sixfold less CAT activity in cells cotransfected with pRSV-TAT than did the corresponding pMCE 69.1 derivative in which the MFNLP was deleted (Fig. 7B). These results support the notion that the MFNLP may be selected in vivo (1, 22, 29, 41, 42, 50, 72, 73) because it confers a repressive effect on transcription (see Discussion).

FIG. 7.

MFNLPs repress transcription in cells where RBF-2 expression has been detected. CAT activity was assayed 48 h posttransfection. Relative CAT activity is indicated on the vertical axis. Each experiment was performed at least twice in triplicate. Error bars indicate the standard deviation from the mean. (A) RBF-2-expressing Jurkat cells were transfected with 4 μg of the indicated pMCE HIV-1 LTR-CAT plasmid (22). Solid bars indicate CAT activity generated by the parent pMCE plasmid, and open bars indicate activity of the pMCE derivatives in which the MFNLP has been deleted by oligonucleotide mutagenesis. (B) RBF-2-expressing Jurkat cells were transfected with the parent clone pMCE 69.1 or the corresponding MFNLP-deleted isogenic construct pMCE 5.13A, as for panel A (solid bars), or cotransfected with pRSV-TAT expressing HIV-1 Tat protein (open bars) or with a corresponding control empty expression plasmid (shaded bars). (C) Undifferentiated HL-60 cells lacking detectable levels of RBF-2 (black and white bars) or HL-60 cells differentiated to monocytes (by treatment with PMA) that express detectable levels of RBF-2 (dotted bars) were transfected with the indicated parent pMCE HIV-1 LTR-CAT plasmid (+MFNLP) or a pMCE derivative deleted of the MFNLP (−MFNLP) as for panel A.

We have previously demonstrated that Jurkat cells express detectable levels of RBF-2 (6). Because our experiments suggest that the MFNLPs provide an additional binding site for RBF-2, we sought to examine the effect of the MFNLP on HIV-1 transcription in a cell line which does not produce RBF-2. We searched for such cell lines, using EMSA and Southwestern blotting. From these experiments we found that various T (Jurkat and CEM), B (Daudi and Ramos), and monocyte (U937) cell lines contained RBF-1 and -2 (results not shown). However, the HL-60 promonocytic leukemia cell line was found to lack detectable levels of both RBF-1 and RBF-2 as demonstrated by EMSA and Southwestern blotting with RBE IV and RBE III oligonucleotides (reference 6 and not shown). Importantly, differentiation of HL-60 cells into macrophages by treatment with phorbol myristate acetate (PMA) was observed to induce the presence of both RBF-1 and RBF-2 (not shown), as detected by both EMSA and Southwestern blotting. To examine whether the presence of RBF-2 contributes to the repressive effect of the MFNLP on LTR transcription, we compared expression of the LTR derivatives in differentiated and undifferentiated HL-60 cells. We found that in undifferentiated cells, where RBF-2 is absent, deletion of the MFNLPs had no effect on transcription of transiently transfected LTR clones pMCE 9.104 and pMCE 69.1 (Fig. 7C). However, in contrast, deletion of the MFNLP from both pMCE 9.104 and pMCE 69.1 caused significantly elevated transcription in differentiated HL-60 cells which contain RBF-2 (Fig. 7C), similar to the effect of the MFNLP that we observed in Jurkat cells (Fig. 7A). Therefore, these results demonstrate that the MFNLP in LTRs from two different patients confers a repressive effect on transcription in cells expressing RBF-2 but not in cells lacking detectable levels of RBF-2.

RBF-1, but not RBF-2, is recognized by antibodies against the Ets family member GABP.

We previously identified RBF-1 and RBF-2 as factors which bound at least four elements within the HIV-1 LTR that are necessary for full transcriptional responsiveness to v-Ha-Ras (6). RBF-1 binds with similar specificity as Ets family members but appears to be immunologically unrelated to c-Ets-1, Fli-1, ERF, or Elf-1 (6). We have found that binding of RBF-1 to RBE IV can be prevented by antibodies against GABPα (Fig. 8A, lane 2) or GABPβ1 (lane 3) but not GABPβ2 (lane 4), suggesting that RBF-1 may contain GABP subunits α and β1 or closely related subunits (68). However, previous data suggest that RBF-1 may contain a DNA-binding subunit of approximately 100 kDa (6), which is significantly larger than GABPα, whose molecular mass is known to be approximately 60 kDa (68). Therefore RBF-1 may be GABPα/β1 and an additional subunit. Alternatively, RBF-1 may represent a differentially produced or posttranslationally modified form of GABP or may be an immunologically related protein that is distinct from GABP.

FIG. 8.

RBF-1 but not RBF-2 contains the Ets family member GABP. (A) EMSA reactions were performed with 1 pmol of radiolabeled RBE IV oligonucleotide and 4 μg of Jurkat nuclear extract. The reactions contained no antibody (lane 1) or 1 μl of rabbit polyclonal antibody raised against GABPα (lane 2), GABPβ1 (lane 3), or GABPβ2 (lane 4). The specific RBF-1 complex is indicated. (B) EMSA reactions were performed with 1 pmol of radiolabeled RBE III oligonucleotide and 4 μg of Jurkat nuclear extract. The reactions contained no antibody (lane 1) or 1 μl of rabbit polyclonal antibody raised against ERF (lane 2), GABPα (lane 3), GABPβ1 (lane 4), or GABPβ2 (lane 5). The specific RBF-2 complex is indicated. NS, nonspecific complex.

We also observe several physical and biochemical similarities between RBF-1 and RBF-2, including their patterns of expression and the apparent sizes of their DNA-binding components (6). Despite these similarities, we find that binding of RBF-2 to RBE III is unaffected by antibodies against GABP (Fig. 8B, lanes 3 to 5). We have also found that RBF-2 is not recognized by antibodies against AP-4 (not shown), despite the fact that RBE III, and sequences duplicated by the MFNLPs, resemble binding sites for AP-4. Furthermore, binding of RBF-2 to RBE III cannot be competed with a strong consensus AP-4 oligonucleotide (Fig. 5, lane 5). We conclude that RBF-2, a factor which specifically binds an element duplicated by the MFNLP, is unrelated to AP-4 or GABP, although it may have a component in common with RBF-1 (6).

DISCUSSION

Retrovirus LTR polymorphism can influence the course of disease. For example, LTRs from murine leukemia virus are major determinants of replication rate, tropism, and pathogenesis (16, 33). Similarly, a 21-bp triplicated motif insertion of feline leukemia virus LTRs causes non-T-cell, non-B-cell spleen lymphomas, increases viral replication, and is postulated to mediate insertional up-regulation of cellular genes from the 3′ LTR (3). For HIV-1 there appears to be no temporal pattern of enrichment for specific LTR polymorphisms (14, 15, 50) or correlation between disease state and specific LTR polymorphisms (22) even though the 3′ LTR is potentially transcriptionally active (40) and despite the high replication rate of HIV-1 in vivo (11, 31).

Several groups have now reported the occurrence of a partial TCF-1α duplication, CTG motif or MFNLP, 5′ of the NF-κB enhancer elements in HIV-1 LTRs from a significant number of patients, living in geographically distinct areas of the world, sampled over different years (1, 22, 29, 39, 41, 42, 50, 72, 73). This polymorphism is also found in the full length ANT-70 molecular clone (52). The in vivo prevalence of this polymorphism argues for an important role in the HIV-1 life cycle, despite a lack of correlation between its occurrence and disease state (22). Interestingly, we have found that the presence of the MFNLP correlates with the occurrence of mutations to the binding sites for RBF-1 and RBF-2 (6, 22). In LTRs harboring an MFNLP, 86% have mutations to RBE IV, III, or II sites (22). Thus, we have previously suggested that the MFNLP may provide a compensatory function rather than a novel pathogenic effect, in contrast to what has been observed for some onconeogenic retrovirus LTR polymorphisms of other viruses (3, 16, 33).

Our compensatory hypothesis is supported by analysis of LTR sequences containing MFNLPs reported by other groups. Specifically, the presence of the CTG motif duplication reported by Golub and coworkers occurred in the context of an RBE II site mutation (29). Furthermore, several of the LTRs possessing partial TCF-1α duplications (3B-5 and 3B-7), reported in a longitudinal study, had mutations to the RBE IV site (50) that ablate binding of RBF-1 (6). A similar partial TCF-1α duplication was reported to be present in a fully pathogenic and replication-competent variant that completely lacked the RBE II site and the overlapping NF-κB and Ets sites (73). The dispensability of the NF-κB binding sites for induction of AIDS by HIV-1 is paralleled by a similar finding for simian immunodeficiency virus (35). In contrast to the dispensability of NF-κB for frank progression to AIDS, it is interesting that due to the occurrence of the MFNLP duplication, the RBE III site is 100% conserved in all available HIV-1 LTR sequences except those reported from a long-term nonprogressor with stable CD4 counts (13).

Our results indicate that the MFNLPs bind a specific nuclear factor that appears to be identical to RBF-2 (6), which is consistent with the fact that these polymorphisms represent a duplication of RBE III. Although previous reports have described the MFNLP as a partial TCF-1α/hLEF duplication (50, 72, 73), we find that most MFNLPs do not bind hLEF in vitro. Furthermore, taken together with our previous analysis of 500 HIV-1 LTR sequences from patients, our results suggest that the high-affinity binding site for hLEF which is present within the prototype HXB2 LTR is not well conserved in vivo (22). However, we cannot exclude the possibility that the lower-affinity hLEF sites that we observe on many MFNLPs and upstream sites may be involved in cooperative interactions of hLEF with other factors bound to the LTR in vivo, as shown in vitro for hLEF, TFE3, Ets, and NF-κB (62).

At least some MFNLPs contain a core GGA(A/T) sequence, which can be bound by recombinant Ets protein in vitro. However, MFNLPs which contain an Ets binding site appear to be an exception rather than a conserved feature of these duplications. We have also observed that some naturally occurring LTRs appear to have Ets binding sites within the LTR/Nef-coding region and that several of these are a result of deletions which create a novel GGA core sequence (Fig. 6A). The significance of these novel Ets binding sites on HIV-1 transcription remains to be determined.

The identity of RBF-2 also remains to be determined. Because neither a competitor oligonucleotide containing a strong consensus AP-4 binding site nor antibodies against AP-4 protein affect RBF-2 binding, we do not believe these factors could be the same. In UV cross-linking experiments, Koken and coworkers have detected a 68-kDa band that interacts with the CTG motif (42). By Southwestern blotting and UV cross-linking, we find that the DNA-binding component of RBF-2 appears to have a molecular mass of 100 kDa (6). We note that the CTG motif-interacting proteins detected by Koken and coworkers migrate as multiple species between 68 and 100 kDa, indicating that these experiments may also have detected RBF-2 (42).

Despite differing in DNA-binding specificity, RBF-2 is similar to RBF-1 in its pattern of expression, the apparent molecular weight of a DNA-binding component, its mobility in native gels as a complex with DNA, and its requirement for v-Ras and protein-tyrosine kinase responsiveness of the HIV-1 LTR. RBF-1 and RBF-2 were also found to elute identically in heparin-agarose-fractionated nuclear extracts (not shown). Furthermore, we find that the common 100-kDa species detected by Southwestern blotting with RBE III and RBE IV oligonucleotides generates identically sized fragments in partial proteolysis experiments (not shown). Taken together, these results suggest that RBF-1 and RBF-2 may have a common 100-kDa subunit. We demonstrate here that RBF-1 likely contains GABP β1, subunits α and β1, whereas RBF-2 does not. GABPα is a 60-kDa protein containing a C-terminal Ets domain which requires GABPβ for DNA-binding specificity (65, 68, 69). The finding that RBF-1 contains GABP subunits is consistent with the identified significance of RBE IV and III for v-Ras responsiveness (6) and the fact that RBF-1’s DNA-binding specificity is identical to that of Ets proteins (6). GABP has previously been demonstrated to bind to the Ets sites within the NF-κB motifs and contribute to v-Raf responsiveness (24). Ets family members are known to form complexes with a variety of different factors (4, 5). Complexes between c-Ets and NF-κB as well as NF-AT have been shown to be involved in the activation of the HIV-1 LTR (4). Based on these observations, we suggest that RBF-1 may represent a complex of GABPα/β1 and an additional 100-kDa subunit which is similar or shared with RBF-2.

Our results, taken together with the fact that the presence of the MFNLP correlates with mutations to RBE sites, suggest that the MFNLP is selected in vivo because it provides a target for RBF-2 binding that is essential to HIV-1 replication. In cells where RBF-2 DNA-binding activity is detectable, we find that the MFNLPs mediate a repressive effect, whereas in contrast the MFNLPs have little effect in cells that do not express RBF-2. Therefore, it is possible that the MFNLP is selected in vivo because it down-regulates HIV-1 transcription during monocyte to macrophage differentiation or during T-cell activation. For example, a small reduction in the temporal expression of HIV-specific proteins could drastically influence the survival of a newly infected cell in the face of coinfiltration of HIV-specific cytotoxic T lymphocytes into splenic white pulp (9, 10). An additional possibility is that a more significant effect of the MFNLP is masked in our experiments because we have used transiently transfected LTR templates. It is possible that the MFNLP alters an aspect of HIV-1 transcription in the context of chromatin. This would be consistent with linker scanning mutations within the RBE III and IV regions (without the benefit of the MFNLP) that have been found to drastically impair HIV-1 viral replication (71). The recent finding that a fully pathogenic HIV-1 isolate, capable of inducing CD4 decline, clinical deterioration and AIDS, completely lacks NFκB, RBEII, and Ets enhancer sites but has an MFNLP that duplicates only the RBE III/RBF-2 site (73), also strengthens the notion that MFNLPs may compensate for impaired binding sites found elsewhere in the LTR. We believe these issues will be more clearly resolved upon determining the molecular composition of RBF-2.