INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Foamy viruses, including human foamy virus (HFV) and the simian foamy viruses (SFVs), are unusual among retroviruses in that they contain two distinct promoter elements (25). As in all retroviruses, a promoter element located in the viral long terminal repeat (LTR) directs the synthesis of a genome-length transcript that gives rise to the viral structural proteins Gag, Pol, and Env. The second foamy virus promoter, termed the internal promoter (IP), is located at the 3' end of the envelope gene (4, 16, 19). The IP is primarily responsible for the expression of two open reading frames located between env and the 3' LTR, one of which encodes a transcriptional transactivator termed Bel-1 in HFV and Tas in the SFVs (4, 15). Importantly, both of these promoter elements are strongly activated upon expression of the cognate Bel-1 or Tas regulatory protein (4, 14, 15, 21, 26, 28). Mutational analysis has demonstrated that IP function is critical for foamy virus replication in culture, most probably because loss of IP function results in a marked drop in the expression of Bel-1/Tas, which is known to be essential for viral replication (17, 18).

Analysis of the mechanism of action of Bel-1 in HFV and of Tas in SFV-1 has demonstrated that these virally encoded transcription factors are DNA binding proteins, a property which distinguishes Bel-1/Tas from the functionally equivalent human immunodeficiency virus type 1 Tat and human T-cell leukemia virus type 1 Tax proteins (12, 30). Binding sites for Bel-1 have been defined in both the IP and LTR promoter of HFV (12, 13), while binding sites for Tas have been identified in the SFV-1 IP and in the gag gene (3, 30), although the role and importance of this latter Tas-dependent enhancer element remains unclear. For HFV, it has been demonstrated that the Bel-1 DNA binding site located in the IP displays a far higher affinity for Bel-1 than the major Bel-1 DNA binding site located in the LTR promoter (13). It has been hypothesized that this difference in affinity may, at least in part, explain the observation that the HFV IP is activated significantly earlier than the LTR promoter during the foamy virus life cycle (15, 19). This difference in affinity, if functionally important, could also explain the limited homology noted between the HFV LTR and IP Bel-1 binding sites (13). Similarly, the SFV-1 IP and Gag gene Tas binding sites also display only limited homology to each other and to Tas-responsive DNA sequences present in the SFV-1 LTR (3, 20, 30). These findings, combined with the observation that Bel-1 and Tas are specific only for target sequences present in their cognate viral genome (4, 13), have meant that no consensus DNA binding site for either Bel-1 or Tas has been proposed, even though minimal DNA binding sites for both proteins have been defined (3, 12, 13, 30).

In this study, we have used a novel in vivo randomization and selection procedure to define a consensus DNA binding site for the SFV-1 Tas protein. We demonstrate that this consensus DNA sequence binds Tas with a higher affinity than does the wild-type IP Tas binding site both in vivo and in vitro. In addition, we have constructed a highly variant Tas DNA binding site that, while identical to the minimal IP Tas binding site at only 8 of 25 positions, can nevertheless respond effectively to Tas in vivo. The flexibility of the DNA binding specificity of Tas appears to explain the marked variability in the sequence of Tas responsive sequences previously identified in the SFV-1 genome.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Construction of molecular clones. The yeast reporter plasmids pLGSD5 and pJLB have been described previously (8, 10). Both contain the lacZ indicator gene under the control of a basal yeast cyc promoter. However, pJLB differs from pLGSD5 in that the former lacks the yeast cyc promoter regulatory element UAS_G, which contains negative regulatory elements (8). As a result, the pJLB vector displays a higher basal activity but also a significantly stronger response to weak transactivators. A pJLB-based yeast indicator plasmid containing the minimal Tas DNA binding site (69 to 45) from the SFV-1 IP has been described previously (13). This wild-type Tas binding sequence, the minimal Bel-1 binding site (163 to 139) defined in the HFV IP, and various mutant Tas DNA binding sites (UP, VA, HS1, and HS2 [see Fig. 3]), were synthesized as oligonucleotides with flanking XhoI sites, annealed to make them double stranded, and then cloned into the XhoI site 5' proximal to the cyc promoter present in the indicator plasmids pLGSD5 and pJLB. In addition, wild-type and defective (knockout [KO]) forms of a larger SFV-1 IP sequence, extending from 150 to 29 relative to the IP transcription start site, were PCR amplified with primers that introduced flanking XhoI restriction sites, using the corresponding mammalian expression constructs (described below) as templates, and cloned into the XhoI site of pJLB. Orientation of inserts was screened by PCR and confirmed by DNA sequencing.

In vivo randomization selection of Tas response sequences. The randomized Tas binding site libraries R1 to R5, as well as R3A and R3B, were generated by a single-step PCR procedure. The 5' primers used were as follows: 5'-GACCGCTCGAG TT GCA ATCAC TGGAA ATAGA AGTTA CAGATCCGCCAGGCGTGT-3' R1  R2 R3 R4 R5

Yeast transformation and analysis. The lacZ indicator plasmids containing various mutant Tas DNA target sequences and the single-copy, pYCplacIII-based Bel-1 or Tas yeast expression plasmid (9, 13) were cotransformed into the yeast strain PSY316. After 3 days of growth selection on Ura Leu plates, overnight cultures were prepared and cell extracts were assayed for -gal activity as described previously (1).

Mammalian cell culture and transfection. COS cell cultures (35-mm plates) were transfected by the DEAE-dextran procedure (6) with 1 µg of a CAT-based indicator construct containing wild-type or mutant forms of the SFV-1 IP, and up to 200 ng of the pcTas expression plasmid. To maintain a level of 1.2 µg of DNA per transfection in all experiments, transfection cocktails were also supplemented with the parental, negative control plasmid pBC12/CMV where necessary. Induced CAT activities were assayed at ~50 h posttransfection as previously described (22).

Gel retardation analysis. The fusion protein glutathione S-transferase (GST)-Tas was expressed in the protease-deficient E. coli XA90 and purified as previously described (13, 30). The SFV-1 IP DNA sequences used for gel retardation and competition analysis were generated by PCR and extended from 159 to +13 in the SFV-1 IP. The wild-type DNA probe was labeled with [-³²P]ATP with T4 polynucleotide kinase, and the total isotope incorporation was determined by scintillation counting after column purification. The binding reaction was carried out with ~10⁴ cpm (~0.2 ng) of the probe and 50 ng of GST-Tas fusion protein in 40 µl of binding buffer (10 mM Tris-Cl [pH 7.5], 65 mM KCl, 2.5 mM MgCl₂ 1 mM dithiothreitol, 5% glycerol) containing 200 ng of poly(dI-dC) and 40 µg of bovine serum albumin. Binding was allowed to proceed for 30 min at 4°C, and the reaction products were resolved on a 5% native polyacrylamide gel and visualized by autoradiography. Competitor DNA fragments were PCR amplified from the respective cat-based mammalian indicator plasmids described above. A 154-bp DNA fragment containing the Mason-Pfizer monkey virus constitutive transport element (7) was amplified by PCR and served as a nonspecific competitor DNA. For competition experiments, competitor DNAs were added at an 8- or 80-fold molar excess over the labeled probe fragment and were incubated with GST-Tas for 10 min before addition of the probe. The results of competition experiments were quantitated with a PhosphorImager and Image QuaNT software (Molecular Dynamics).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Previously, Zou and Luciw (30) defined a 25-bp Tas DNA binding site in the SFV-1 IP by using DNase I protection analysis and also showed by gel shift analysis that this DNA sequence was sufficient for specific Tas binding in vitro. Subsequently, we demonstrated that a single copy of this 25-bp sequence, which extends from 69 to 45 in the SFV-1 IP relative to the transcription start site, is sufficient to render a yeast cyc promoter element highly responsive to Tas when introduced 5' to this minimal promoter and analyzed in yeast cells (13).

To define the consensus sequence for Tas DNA binding in yeast cells in vivo, we subdivided the minimal 25-bp SFV-1 IP Tas DNA binding site into five regions of 5 bp each, termed R1 through R5, as shown in Fig. 1. Each 5-bp segment was then individually randomized in the context of the lacZ-based indicator construct pLGSD5 to give five libraries, each containing 1,024 different candidate Tas target sequences (see Materials and Methods for details). Each library was cloned into E. coli and analyzed for integrity before being introduced into the yeast strain PSY316 together with a yeast Tas expression plasmid. After 3 days of selection, yeast colonies containing DNA sequences that retained Tas responsiveness were identified by staining with the -gal substrate CPRG. Approximately 4,000 to 5,000 colonies were screened per library. Individual responsive colonies were then picked, and the responsive plasmids were recovered. The Tas responsiveness of each indicator plasmid was validated by retransformation into yeast in the presence and absence of a Tas expression plasmid, and the precise level of -gal activity induced by Tas, for each responsive indicator plasmid, was quantified as previously described. As shown in Fig. 1, a global analysis of the five libraries obtained during this research effort demonstrated that the central R3 library contained very few strongly positive clones (~0.5%). A higher but still modest percentage of positive colonies was obtained with the closely flanking libraries R2 and R4 (~5 and ~2%, respectively). In contrast, the R1 and R5 libraries, representing the 5' and 3' extremes of the SFV-1 IP Tas binding site, displayed high levels of positivity, with ~70 and ~20%, respectively, showing readily detectable -gal activity.

View larger version (59K): [in this window] [in a new window]   FIG. 1.   Selection of Tas-responsive sequences from randomized libraries in vivo. Introduction of a 25-bp Tas binding sequence, derived from the SFV-1 IP, adjacent to a minimal yeast cyc promoter element renders this promoter highly responsive to the Tas transactivator. To identify residues required for activation by Tas, this 25-bp sequence was subdivided into five regions of 5 bp each, termed R1 through R5, and each 5-bp sequence was independently randomized. Tas-responsive (rows 1 to 12) and Tas-nonresponsive (rows 13 and 14) DNA sequences were then selected after transformation into yeast cells and sequenced. Recovered strongly responsive sequences for each region R1 to R5 are aligned with the SFV-1 IP wild-type sequence (row w.t.). The first row also gives the approximate overall percentage of highly Tas responsive clones observed in each library by staining of yeast transformants on filters. The R3 library was further subdivided into an R3A library, involving randomization of only the 5' two residues of R3, and an R3B library, involving randomization of the 3' three residues. Residues held constant in these libraries are shown as lowercase letters in the upper section of column R3. Some sequences were recovered more than once, as indicated by a number in parentheses next to the relevant sequence. Rows 13 and 14 give Tas-nonresponsive sequences recovered during this screen, with residues not observed in any recovered positive sequence indicated in lowercase letters. The level of induction mediated by each selected sequence in yeast cells upon coexpression of Tas is given to the right of each sequence, as a multiple of the activity seen with the wild-type IP sequence, which is arbitrarily set at 1.0.

The small number of strongly positive clones observed with the R3 library, all of which displayed the wild-type sequence, prompted us to generate two additional libraries, termed R3A and R3B, in which only the 5' 2 bp or the 3' 3 bp of the R3 sequence were randomized. This allowed us to identify additional, non-wild-type sequences covering this region that retained at least partial Tas responsiveness. Such clones were difficult to identify in the original R3 library since this particular library generated a significant background of transcriptionally active but Tas-nonresponsive plasmids (data not shown).

Consensus Tas DNA binding sequence. An analysis of the Tas-responsive sequences obtained in the above screen is shown in Fig. 1, and the data are summarized in Fig. 2, which also presents the derived consensus Tas binding sequence. Based on these data, it is apparent that the core 5'-TGGAA-3' sequence covered by the R3 library is critical for Tas binding. Three of these positions were found to be invariant in this analysis, while variation at the other two positions was infrequent and, when observed, resulted in a marked drop in Tas responsiveness.

View larger version (13K): [in this window] [in a new window]   FIG. 2.   Consensus Tas DNA binding sequence. This figure presents a compilation of the sequence data presented in Fig. 1, with the frequency of recovery of each base at each location in the randomized Tas-responsive sequence given as a percentage. The most prevalent base at each position is highlighted. At the bottom, these data are summarized to give a consensus Tas DNA binding site. The following abbreviations are used: W = A or T; Y = C or T; S = G or C; K = G or T; and N = largely random.

Biological activity of Tas binding site variants in yeast. The data presented in Fig. 1 and 2 suggest that it should be possible to derive highly variant forms of the Tas IP DNA binding site that retain good, or even enhanced, Tas binding ability. To test whether this was indeed the case, we designed two 25-bp oligonucleotides termed UP and VA (for variant). The UP sequence differs at seven positions from the wild-type SFV-1 IP Tas binding site (Fig. 3A) and contains the optimal nucleotide sequence for Tas binding at every position, as predicted from the data shown in Fig. 1. In contrast, the 25-bp VA sequence was designed to be maximally different from the wild-type SFV-1 IP sequence yet to still largely conform to the Tas DNA binding site consensus given in Fig. 2. It was therefore anticipated that the VA sequence, which differs from the wild-type sequence at 17 of 25 positions (Fig. 3A), would retain detectable but not maximal Tas binding ability.

View larger version (26K): [in this window] [in a new window]   FIG. 3.   Tas DNA binding site variants. This figure gives the sequence of the UP, VA, and KO mutants of the 69 to 45 SFV-1 IP minimal Tas DNA binding site and of the HS1 and HS2 mutants of the 163 to 139 HFV IP minimal Bel-1 DNA binding site.

Biological activity of Tas binding site variants in mammalian cells. All the in vivo data on Tas DNA binding presented thus far were obtained with yeast cells, and we therefore wished to confirm these findings in the more physiologically relevant context of cultured simian cells. Previous mutational analysis of the SFV-1 IP (4, 30) has suggested that this promoter contains only three sequence elements required for maximal Tas response (Fig. 4A). These are a TATA box, seen in the majority of RNA polymerase II-specific promoters; the Tas binding site, located between residues 45 and 69 relative to the transcription start site; and a third DNA sequence, termed the distal element, located between nucleotides 108 and 140 (30). While the distal element remains only poorly characterized, it does not bind Tas (30) (Table 1) and is therefore thought to serve as a cellular factor binding site. Surprisingly, previous work (4, 30) has demonstrated that mutation of either the distal element or the Tas binding site sequence alone results in only a two- to fourfold drop in the Tas responsiveness of the IP; however, mutation of both sites results in the loss of Tas response (30). The observation that the single DNA target site for Tas present in the SFV-1 IP can be mutated with only a partial loss of Tas responsiveness is surprising and prompted us to reexamine the biological activity of the SFV-1 IP in mammalian cells and, in particular, to test the ability of wild-type and mutant SFV-1 IP derivatives to respond to a range of different Tas expression levels.

View larger version (14K): [in this window] [in a new window]   FIG. 4.   Activity of the SFV-1 IP in mammalian cells. (A) Structure of the SFV-1 IP. Functionally important elements include a TATA box, the Tas DNA binding site (TBS) located between 69 and 45 relative to the transcription start site (arrow), and a distal element (DE), located between 108 and 140, that is believed to form a cellular Tas cofactor binding site. (B) Dose-response curves of SFV-1 IP derivatives with Tas in transfected COS cells. Indicator constructs, consisting of the indicated SFV-1 IP derivatives linked to the cat indicator gene, were transfected into COS cells together with increasing levels of the Tas expression plasmid pcTas, as indicated. The cells were transfected with 1 µg of the relevant indicator construct and up to 200 ng of pcTas. A DNA level of 1.2 µg per transfection was maintained by supplementation with the parental pBC12/CMV plasmid, which also serves as a negative control. Transfected cells were harvested ~50 h after transfection, and CAT levels were determined as previously described (22).

View larger version (36K): [in this window] [in a new window]   FIG. 5.   Activation of SFV-1 IP mutants by Tas. Indicator constructs consisting of the SFV-1 IP (159 to +13 or 86 to +13) linked to the cat indicator gene were mutated by introduction of the UP, VA, or KO mutations (Fig. 3) into the Tas DNA binding site. The resultant constructs were then transfected into COS cells in the presence of limiting (8 ng) or saturating (200 ng) levels of the pcTas indicator construct. The pBC12/CMV plasmid served as a negative control. Induced CAT expression levels were quantified at ~50 h posttransfection. These data represent the mean for three experiments, with standard deviations indicated by error bars.

Tas DNA binding specificity in vitro. As a final confirmation of the validity of the Tas binding consensus given in Fig. 2, we examined the ability of the UP, VA, and KO mutants of this SFV-1 sequence to compete with the wild-type SFV-1 IP for binding to recombinant Tas protein by using a competitive electrophoretic mobility shift assay. As shown in Fig. 6, addition of recombinant GST-Tas to a labeled SFV-1 IP DNA probe resulted in the detection of two specific DNA-protein complexes labeled C1 and C2. The origin of the C2 complex is unclear, but this could reflect dimerization of either the Tas protein or GST, which is known to form dimers. Formation of both these complexes is effectively competed by an excess of the unlabeled wild-type SFV-1 IP DNA sequence (lanes 3 and 4) but not by an equivalent level of a nonspecific competitor DNA (lanes 11 and 12). Importantly, the addition of an SFV-1 IP-derived DNA sequence bearing the UP mutation resulted in a significantly enhanced level of competition (lanes 5 and 6), demonstrating that the UP mutation indeed enhances Tas binding in vitro. In contrast, an SFV-1 IP DNA competitor bearing the KO mutation did not compete for binding any more effectively than the nonspecific DNA competitor did (compare lanes 9 and 10 with lanes 11 and 12). Finally, a sequence containing the VA mutation of the SFV-1 IP also competed for Tas DNA binding in vitro, although perhaps slightly less efficiently than the wild-type IP sequence did (lanes 7 and 8). This result appears to mirror the in vivo biological activity of the VA sequence observed in yeast cells (Table 1), which was also slightly lower than that seen with the wild-type IP sequence. In conclusion, these in vitro Tas binding data are therefore in agreement with the in vivo data derived from the yeast (Table 1) and mammalian (Fig. 5) systems and serve to further validate the Tas DNA binding consensus given in Fig. 2.

View larger version (61K): [in this window] [in a new window]   FIG. 6.   DNA binding by SFV-1 Tas in vitro. The end-labeled DNA probe used in this EMSA consists of the wild-type 159 to +13 SFV-1 IP sequence. Incubation with recombinant GST-Tas (50 ng) resulted in the detection of two retarded DNA-protein complexes labeled C1 and C2. Preincubation with an 8- or 80-fold excess of the unlabeled wild-type (WT) 159 to +13 DNA fragment inhibits complex formation (lanes 3 and 4), while the same level of a nonspecific (NS) competitor has little or no effect. Competition by the 159 to +13 IP DNA sequence bearing the UP, VA, or KO mutation of the Tas binding site (Fig. 3) is also shown. The percent residual Tas DNA binding, compared to the binding with no added competitor (lane 2), was determined by using a PhosphorImager and is given at the bottom of the figure. Free Int. Pr. Probe, free internal promoter DNA probe.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The replication of primate foamy viruses is dependent on the biological activity of not only the viral LTR promoter but also a second promoter, the IP, located toward the 3' end of the env gene (4, 16, 19). Both promoter elements are strongly activated by the virally encoded Bel-1/Tas transcriptional transactivator (4, 14, 15, 21, 26, 28), and it is therefore the interplay of Bel-1/Tas with these two promoter elements that largely controls the level of foamy virus gene expression.

Analysis of the biological activity of HFV Bel-1 and of SFV-1 Tas has demonstrated that these viral transcription factors are DNA binding proteins and has identified minimal, 25-bp binding sites for Bel-1 in the HFV LTR and IP and for Tas in the SFV-1 IP and gag gene (3, 12, 13, 30). However, comparison of these sequences with each other and with those of other phenotypically defined Bel-1 or Tas response elements has failed to identify a clear DNA binding consensus. The lack of such a consensus reflects two findings. First, Bel-1 and Tas interact only with sites present in their cognate virus (4, 13), thus making any comparison of the DNA binding sites of these two related proteins uninformative at present. Second, it appears that the foamy virus IP has evolved to contain a significantly higher-affinity target sequence for Bel-1/Tas than has the LTR promoter (13), thus suggesting that Bel-1/Tas binding sites in the LTR are likely to diverge significantly from the optimal Bel-1/Tas DNA binding consensus.

In this paper, we report the use of a novel randomization and selection strategy with yeast cells to define the DNA binding consensus for SFV-1 Tas. Yeast cells were chosen because these lower eukaryotes were previously shown to be permissive for transcriptional activation by both Bel-1 and Tas via their respective, mutationally defined DNA binding sites (1, 12, 13). We chose the SFV-1 IP Tas DNA binding site for this analysis because Tas gives a significantly higher level of transcriptional activation than Bel-1 in yeast cells (13), thus facilitating the identification of positive clones, and because the IP, as noted above, contains a particularly high-affinity DNA binding site for Tas (13). An alternative strategy, used for several DNA binding proteins (2, 23, 24, 29), would have been to identify a consensus Tas DNA binding site in vitro by using several cycles of PCR amplification and selection for Tas binding. An advantage of this latter, in vitro technique is that it would have permitted the simultaneous randomization of essentially the entire Tas binding site. In contrast, the in vivo randomization/selection strategy described here is not able to analyze more than five or six nucleotide positions simultaneously because (i) more extensive randomization would require screening an unwieldy number of yeast transformants and (ii) we have found that the percentage of false (i.e., Tas-independent)-positive clones, which largely represent yeast transcription factor binding sites, increases when more nucleotides are randomized. The fact that in any given random library, much of the DNA binding site must therefore remain fixed could affect the range of positive sequences obtained by this approach. However, this concern is balanced by the anticipation that this in vivo approach may be more likely to identify physiologically relevant DNA-protein interactions.

A series of Tas-responsive and nonresponsive sequences identified using by in vivo randomization and selection approach are listed in Fig. 1 and summarized in Fig. 2, which also gives the derived consensus sequence for Tas DNA binding. Based on these data, we suggest that Tas DNA binding sites consist of a core recognition sequence that approximately coincides with the R3 sequence 5'-TGGAA-3'. Flanking this, on either side, are sequences that are defined largely by the R2 and R4 libraries that, while critical for Tas binding, are nevertheless more tolerant of sequence variation. Finally, while the R5 and, particularly, R1 regions contribute relatively little critical sequence information to the Tas DNA binding site, DNA sequences present in these regions can significantly modulate the affinity of Tas for its target site (Fig. 1). Overall, these data indicate that the consensus DNA target site for Tas is somewhat unusual in terms of both its large size and the relative plasticity of this sequence outside of the core R3 region. While it is interesting to speculate that the large size of this target DNA sequence could reflect binding by a Tas dimer or multimer, we note that the Tas binding consensus (Fig. 2) does not give any evidence of being a direct or inverted repeat sequence. Perhaps the most surprising aspect of the Tas DNA binding consensus given in Fig. 2 is the implication that the wild-type SFV-1 IP Tas DNA binding site is suboptimal at several locations, suggesting that this site has not been selected for maximal Tas binding efficiency. Indeed, several of the recovered mutant sequences shown in Fig. 1 proved to be more effective target sites for Tas when tested in yeast cells.

To validate the relevance of the consensus Tas DNA binding sequence given in Fig. 2, we used this consensus to design a sequence, termed UP, that should be fully optimal for Tas binding and a second sequence, termed VA, that should be able to bind Tas effectively despite differing at 17 of 25 positions from the wild-type SFV-1 IP sequence (Fig. 3A). Analysis of the UP and VA sequences in yeast cells (Table 1) and in vitro (Fig. 6) showed that the UP sequence indeed bound Tas more effectively than did the wild-type IP sequence while the VA sequence also bound Tas but somewhat less well than did the wild type. In contrast, an SFV-1 IP sequence mutated in the Tas binding-site core, termed KO, failed to detectably bind Tas in either assay system.

Demonstration of the relevance of the Tas DNA binding consensus in simian cells was complicated by the fact that the SFV-1 IP contains two partially redundant Tas response elements (Fig. 4A). While the first of these coincides with the Tas DNA binding site, the second sequence, referred to as the distal element, fails to bind Tas both in vitro (30) and in vivo (Table 1). This finding, combined with the fact that the distal-element sequence can mediate Tas function in mammalian cells (30) (Fig. 4B) but not yeast cells (Table 1), strongly suggests that the distal element represents a DNA binding site for a cellular cofactor(s). Despite this complication, it nevertheless proved possible to also validate the Tas binding consensus in mammalian cells. In particular, in the context of a minimal SFV-1 IP promoter lacking the distal-element sequence, the KO mutation was found to block activation by Tas in mammalian cells whereas the UP mutation was found to increase the Tas response by two- to threefold. The maximally variant VA mutation was also observed to retain Tas responsiveness (Fig. 5). Together, these data demonstrate that Tas DNA binding in mammalian cells (Fig. 5), as in yeast cells (Table 1) and in vitro (Fig. 6), is effectively predicted by the Tas DNA binding consensus given in Fig. 2.

The data presented in Fig. 5 may also shed some light on the role of the putative cellular distal-element DNA binding factor. The distal-element sequence is not a Tas binding site (30) (Table 1) and has no phenotype in the absence of Tas (Fig. 5), but it can mediate significant activation of the SFV-1 IP by Tas in the absence of an intact Tas DNA binding site (30) (Fig. 4B). This finding suggests that the cellular distal-element binding protein acts by facilitating the recruitment of Tas to the SFV-1 IP. Consistent with this hypothesis, while deletion of the distal element significantly reduces the activation of the SFV-1 IP by Tas, this reduction can be overcome by substitution of the high-affinity UP Tas DNA binding site in place of the wild- type sequence (Fig. 5). In contrast, the UP mutant has only a minimal phenotype in mammalian cells in the context of an SFV-1 IP containing an intact distal element (Fig. 5), suggesting that Tas DNA binding is already highly efficient in this environment. The hypothesis that a cellular cofactor specific for the distal-element sequence significantly enhances Tas binding to the SFV-1 IP in mammalian cells thus may explain why the IP Tas DNA binding site is not more closely homologous to the optimal Tas DNA binding site (Fig. 2). However, it is also possible that the sequence of the IP Tas DNA binding site is simply constrained by the underlying envelope coding sequence.

A final point is that the proposed Tas binding site consensus should permit the identification of Tas binding sites located elsewhere in the SFV-1 genome. Indeed, a comparison of the SFV-1 gag gene Tas DNA binding site (3) with the consensus sequence shows extensive sequence homology (Fig. 7), particularly around the core region. An analysis of regions within the SFV-1 LTR that have previously been implicated in mediating Tas transactivation of the SFV-1 LTR promoter (20) identified several regions with homology to the consensus, including particularly between 1052 and 1028, between 375 and 351, and (in the antisense orientation) between 956 and 980 and between 163 to 187. The question whether these sequences are indeed important for Tas-mediated activation of the SFV-1 LTR promoter can now be experimentally addressed.

View larger version (10K): [in this window] [in a new window]   FIG. 7.   The SFV-1 Tas binding site in gag conforms to the Tas binding consensus. This figure shows an alignment of the Tas binding site in the SFV-1 gag gene, as defined by DNase I protection analysis (3), with the IP Tas binding site and with the Tas binding consensus. A continuous line indicates tight conservation, while a dashed line indicates that this residue was detected in the pool of Tas binding sequences listed in Fig. 1.