INTRODUCTION

Curved DNA structures are sometimes reported to reside within the transcriptional control regions of prokaryotic and eukaryotic genes (1, 2, 3, 4, 5, 6, 7, 8, 9). Various studies have attempted to ascribe a function to curved DNA and extensive data suggesting a role in prokaryotic transcription for the structure has accumulated over the last few years (10, 11, 12, 13, 14, 15, 16, 17, 18). However, little has been revealed as to the role of DNA curvature in eukaryotic transcription. Most of the enhancer elements of the human adenovirus type 2 (Ad2)¹ E1A enhancer are located within a DNA curvature (6). The location of this unusual structure suggests that it may be linked to transcription. This study investigated whether the curved DNA structure stimulated transcription of the E1A gene.

The Ad2 E1A enhancer (also known as the Ad5 enhancer) is comprised of two types of enhancer elements, named element I and II (19). Element I specifically regulates transcription of the E1A gene with repetitions of the element located at positions 200 and 300. Element II regulates transcription in cis across the entire viral chromosome. Four units of element II nest between positions 250 and 280. A curved DNA structure spans the region between positions 200 and 280 (Fig. 1).

Curved DNAs change their conformations at different temperatures: a highly curved DNA conformation usually forms at low temperature, while it is nearly absent at high temperature (20). Based on the temperature-sensitive nature of curved DNA, in vitro transcription assay systems which could distinguish conformational effects of DNA were established. The assay systems provided the first direct evidence that the DNA curvature in the E1A enhancer is involved in the enhancement of transcription. Surprisingly, it was also revealed that a curved DNA of  origin (21) with the proper conformation also stimulated transcription of the E1A gene. Possible roles of the DNA curvature in the function of the enhancer are discussed.

EXPERIMENTAL PROCEDURES

Plasmid ConstructionsAll recombinant DNA methods were performed according to standard protocols (22). The recognition site of each restriction enzyme is indicated using the nucleotide sequence number of the first base pair present at the site.

Construct pP5, which contains the left terminal TaqI and BbeI fragments of Ad2 DNA, was made as follows. The left terminal TaqI fragment (TaqI site; 629 on the viral genome) was obtained by digesting a construct containing the left terminal region of the viral DNA with PstI and TaqI (PstI site had been constructed at the left end of the genome (6)). By digesting the same construct with PstI and BbeI (BbeI site; 813 on the genome), the left terminal BbeI fragment was prepared. The TaqI fragment was cloned between the PstI and AccI sites of pUC19. This construct was named pADT. Then, the BbeI fragment was treated with T4 DNA polymerase, ligated to EcoRI linkers (5-GGAATTCC-3), and digested with EcoRI and AvaI (cleavage site 757). The resulting truncated fragment was gel-purified and cloned between the EcoRI and AvaI sites of pADT to generate pP5.

The following procedure describes construction of pSt. The BbeI fragment of the Ad2 DNA described above was digested with SacII (position 353 on the genome), treated with T4 DNA polymerase, ligated to EcoRI linkers, and finally digested with EcoRI and AvaI. The resulting largest fragment was gel-purified and cloned between the EcoRI and AvaI sites of pADT to generate pSt.

Construct pSt2 is a variant of pSt. It lacked the RsaI site found at position 168 in the pUC19 vector portion of pSt DNA (the position number is for pUC19). The plasmid pUC19 has single ScaI (position 2177) and NdeI (position 183) sites, each of which is also present just once in pSt. The RsaI site is located between these sites. A ScaI and NdeI digest of pSt was treated with T4 DNA polymerase and the larger fragment was isolated from an agarose gel. It was then ligated to a ScaI-RsaI fragment of pUC19 (spanning nucleotides from 2177 to 168) to generate pSt2.

The pSt derivative pCd was prepared by inserting a curved DNA fragment derived from the  origin (spanning nucleotides 38974 ~ 39168 on  DNA) into the EcoRI site of pSt. The fragment had been obtained by digesting  DNA with EcoRI and SspI, and subsequently the SspI end of the fragment had been converted to an EcoRI end by using the EcoRI linker. In pCd, the original EcoRI site of  DNA (position 39168) was ligated to the EcoRI site of the Ad2 DNA and the newly constructed EcoRI site to that of the vector DNA. Sequences of the constructs were verified according to the dideoxy procedure.

Nondenaturing 7.5% polyacrylamide gel (acrylamide to bisacrylamide, 29:1, w/w) electrophoreses were performed in 45 mM Tris borate (pH 8.3) and 1 mM EDTA at 2.4 V cm¹ at 5, 15, 25, 30, and 37 °C. After the electrophoreses, gels were stained with ethidium bromide.

Two types of reactions were carried out: (i) the reactions employing the constructs pP5, pSt, and pCd; and (ii) those employing pSt2.

Transcription reactions for pP5, pSt, and pCd were carried out at 5, 15, 25, and 30 °C. DNA templates were prepared by digesting each construct with the restriction enzymes as follows: BamHI, EcoRI, and PstI for pP5; BamHI, EcoO109I, and PstI for pSt; BamHI, BbeI, and PstI for pCd. After digestion, digests were extracted with phenol, precipitated with ethanol, rinsed with 70% ethanol, and dried. Then, the template DNA solution containing 0.3 µg/µl digest, 9 mM Tris-HCl (pH 7.6), 5 mM NaCl, 1 mM MgCl₂, and 0.4 mM EDTA was prepared. After 15 µl of HeLa cell lysate (BRL), which contained 20 mM Hepes (pH 7.9), 100 mM KCl, 12.5 mM MgCl₂, 0.1 mM EDTA, 2 mM dithiothreitol, and 17% (v/v) glycerol, and 5 µl of template DNA solution were separately preincubated for 5 min at the temperature set for each assay, they were mixed together and further incubated for 5 min at the temperature. Then, 5 µl of NTP solution containing 5 mM creatine phosphate, 2.5 mM each of ATP, CTP, and UTP, 250 µM GTP, 10 µCi of [-³²P]GTP (400 Ci/mmol; Amersham), and 0.7 mM EDTA was added to the above mixture and transcription was initiated. After a 1-h incubation, the transcription was terminated by adding 300 µl of stop solution containing 0.25 M sodium acetate (pH 5.5), 10 mM EDTA, 0.5% SDS, and 1 µg of carrier tRNA and the transcripts were extracted twice with phenol/chloroform (1/1, v/v), and once with chloroform/isoamyl alcohol (24/1, v/v). The transcripts were then precipitated with ethanol, rinsed with 70% ethanol, dried, and resuspended in 10 µl of loading buffer containing 90% (v/v) formamide, 0.01% bromphenol blue, 0.01% xylene cyanol, and 10 mM Tris-HCl (pH 7.6). They were resolved by 4% polyacrylamide, 7 M urea gels (acrylamide to bisacrylamide, 29/1, w/w). Transcription levels were determined using the Fujix BAS2000 Bio-image analyzer (Fuji Photo Film Co., Ltd.).

For reaction ii, the water bath was settled in a cold room set at 5 °C or in an incubation room set at 37 °C. In each room, each transcription reaction was carried out in a micro test tube held in the bath set at 20 °C. The tube, which was not moved throughout an assay, harbored a micro stir bar (6.3 mm³). DNA templates were prepared by digesting pSt2 with EcoO109I, RsaI, AvaI, and PstI. After the digest was treated in the same way as reaction i, template DNA solution containing 3 µg/µl digest, 7.5 mM Tris-HCl (pH 7.6), 7.5 mM NaCl, and 1 mM MgCl₂ was prepared. Prior to addition, the template DNA solution was preincubated for 1 h in each room (5 or 37 °C). Also preincubated for more than 1 h in each room were micropipette tips used for the transfer of the template DNA solution. In addition, 7.2 µl of the HeLa cell lysate and 4.8 µl of a NTP solution were separately preincubated for 5 min in the water bath (20 °C). The NTP solution contained 2.5 mM creatine phosphate, 1.25 mM each of ATP, CTP, and UTP, 125 µM GTP, 10 µCi of [-³²P]GTP (400 Ci/mmol), and 0.35 mM EDTA. In the assay, at first, the HeLa cell lysate (20 °C) and the NTP solution (20 °C) were mixed in the micro test tube containing the micro stir bar and incubated for 5 min at the temperature (20 °C). After the incubation, 0.5 µl of template DNA solution (5 or 37 °C) was added to the mixture and transcriptions were initiated. Stirring was started at 1 min before the addition of the DNA solution and stopped at 1 min after the addition. Reactions were carried out for 10 or 40 min at 20 °C. The termination of the reaction and the analysis of the products were carried out in the same way as reaction i.

RESULTS

Three test templates were used in the first assay. Each of the constructs pP5, pSt, and pCd contained an internal control (reference) template and a test template (Fig. 2). By completely digesting each construct with the appropriate restriction enzymes, each test template and reference template could be excised in precisely equimolar amounts. Therefore, the transcription level of each test template could be precisely determined after normalization to the reference signal. The lengths of the test templates were designed to be as close as possible to avoid any effect of size. The template Wt (wild-type, 783 bp) contained the entire 5-flanking sequence of the Ad2 E1A transcription unit. Template St (straight, 816 bp) lacked the E1A enhancer and also a region further upstream of the enhancer (deleted region, nucleotide positions from 1 to 355), but carried instead a DNA sequence derived from pUC. Template Cd (curved, 763 bp) carried a curved DNA of  origin instead of the E1A enhancer. Templates St and Cd both carried a complete E1A promoter sequence. All the test templates were expected to produce the same transcript of 269 nucleotides in length. The reference template (651 bp) also contained the entire 5-flanking sequence of the E1A gene and was expected to produce a transcript of 147 nucleotides.

Temperature-dependent conformational changes of the E1A enhancer and those of the corresponding regions of the mutants were investigated (Fig. 3). Curved conformations of the fragments Wt and Cd changed significantly in a temperature-dependent manner, while St which contains the EcoO109I-EcoRI region of the template St did not show any electrophoretic anomaly at the test temperatures. This result clearly demonstrated that the EcoO109I-EcoRI region of the St does not contain a curved structure. The fragment Wt was electrophoresed very anomalously at 5 and 15 °C. According to the study of Calladine et al. (23), plane curves are electrophoresed more slowly than space curves. Taking this knowledge into consideration, planarity of the Ad2 curvature seems to be extremely high at low temperature. Electrophoretic anomaly for Cd was not so marked as that for Wt at each temperature. The principal cause of the phenomenon may be ascribed to the position of the curvature in Cd. It is known that a curvature located at the terminal region of a fragment shows less anomaly than that located near the center of the fragment (24).

Using the templates described above, in vitro transcription assays were carried out (Fig. 4A). At 5 and 15 °C, each test template was transcribed faithfully to produce the transcript of 269 nucleotides (lanes 1-6). However, at 25 and 30 °C, unexpected transcripts of a slightly smaller size were also observed (lanes 7-12). They began to emerge at 25 °C and became somewhat dominant at 30 °C. From the reference template, two kinds of transcripts were also produced at 25 and 30 °C. The upper bands corresponded to the expected size of 147 nucleotides. The lower bands were the unexpected ones and they were dominant at 30 °C in this case as well. An experiment using purified templates showed that these unexpected transcripts derived from the transcription templates and not from the vector DNA (data not shown). They may have been transcribed from an unexpected initiation site located only several bases downstream from the authentic initiation site of the E1A gene or they may have been the degradation products of the expected transcripts. These transcripts were included in calculating relative transcription levels (Fig. 4, B and C).

Effects of conformational change in the Ad2 enhancer or in a DNA curvature of  origin on transcription of the E1A gene.

The stimulation of transcription by the E1A enhancer was expressed as the ratio of the transcription level for the template Wt to that for the St (Fig. 4B). At 25 and 30 °C, the mean values of the transcript ratio, Wt/St, were 1.90 and 1.97, respectively, indicating that the E1A enhancer functioned at both temperatures in almost the same way. At 5 and 15 °C, however, values of 1.17 and 1.19 showed that the enhancer hardly functioned. As shown in Fig. 3, the E1A enhancer was moderately curved at 25 and 30 °C, and severely curved at 5 and 15 °C. A correlation exists between the extent of stimulation of transcription and the degree of enhancer bending. The Ad2 genes are expressed in human cells. Therefore, the enhancer conformation around 37 °C should be the best for stimulation of transcription. However, that could not be confirmed (at 37 °C, HeLa cell lysate nearly lost the activity to drive transcription). All reactions were carried out for 1 h. At 5 °C, the amounts of transcripts slightly increased when a reaction was carried out for 4 h (data not shown). However, the longer reactions did not seem to improve the Wt/St value. The quality of the reaction seemed to have changed.

Fig. 4C shows the values of Cd/St. Both templates carry the E1A promoter but do not carry the E1A enhancer. Thus, the transcription levels from both templates were expected to be always the same. The results at 25 and 30 °C met the expectation. The mean values of Cd/St were 0.97 at 25 °C and 1.03 at 30 °C, respectively. However, the results at 15 and 5 °C did not meet the expectation. At 15 °C, the value was 0.82, suggesting that the template Cd behaved relatively negatively in the transcription. The continuous decrease in the Cd/St value from 1.03 to 0.82 correlates well with the gradual change in the  DNA conformation from moderately curved to more pronouncedly curved. Surprisingly, at 5 °C, the upstream region of the template Cd stimulated the transcription of the E1A gene by 1.70 times. This value was very near to the optimum value of 1.97 that was observed at 30 °C for transcriptional stimulation directed by the wild-type enhancer. Even when compared with the wild-type enhancer, this region stimulated transcription at this temperature by about 1.5 times (1.70/1.17).

Although transcription of the E1A gene from templates Cd and St was expected to be equivalent, this was not the case: Cd repressed or enhanced transcription. The only possible explanation for these results is that the conformation of the  curvature influenced promoter function in vitro. The  DNA was located more than 110 bp upstream from the TATA box. A related report describes that a synthetic DNA curvature inserted into about 30 bp upstream from a TATA box was a potent activator. However, in this case, moving the sequence into an additional 55 or 110 bp upstream had little effect on its ability to activate transcription (25). The effects generated by the  curvature are, therefore, the first experimental evidence that an exogenous DNA curvature can influence transcription from a distant site.

To rule out the possibility that something other than the enhancer conformation was affecting transcription in previous experiments (Fig. 4B), a new set of experiments was conducted in which the reaction components other than DNA templates were kept at a constant temperature. Fig. 5A illustrates these experiments. All the components except the DNA templates were mixed in a micro test tube containing a micro stir bar and were preincubated for 5 min at 20 °C. Each tube was held in the water bath and was not moved throughout the assay. After preincubation of the nontemplate-reaction components, the preincubated template DNA solution was added to the mixture. The added DNA represented only 1/36 of the reaction volume including the stir bar, and the DNA addition did not disturb the temperature of the reaction mixture as monitored by a thermocouple (, 0.5 mm).

Transcription templates were prepared by digesting pSt2 as shown in Fig. 5B. The template tWt functioned as a test template and tSt as a control template in the same way that the template St did as shown in Fig. 4 (tSt does not contain any curvature and thus does not show temperature-dependent conformational change). The lengths of tWt and tSt were designed to be as close as possible to avoid any effect of size: tWt with 656 bp and tSt with 675 bp were expected to generate transcripts of 152 and 141 nucleotides, respectively. Comparison of the amounts of both transcripts in a single assay should give a direct estimate of the extent of transcriptional stimulation. Fig. 5C shows the results. In each reaction, the template tWt generated an unexpected transcript with a slightly smaller size, whereas tSt only generated faithful transcript. This result suggested that the unexpected transcript was generated not by degradation but by perturbed transcription. For the transcription of tWt, the amounts of the expected and unexpected transcripts were added together as had been done in the above experiment. Preincubation of the DNA solution at 37 °C (Fig. 5C, lanes 2 and 4) slightly more positively influenced transcription of tWt than did preincubation at 5 °C (lanes 1 and 3). The results were quantitated mechanically (Fig. 5D). When the DNA solution was preincubated at 37 °C, the mean values of the transcripts ratio, tWt/tSt, were higher (1.31 at 10 and 40 min) than when the solution was preincubated at 5 °C (1.07 at 10 min and 1.14 at 40 min). Interestingly, a set of results for 10-min reactions was almost the same as those for 40-min reactions. In this assay system, transcription factors stayed at 20 °C constantly. In addition, DNA conformation of the template tSt was not influenced by the surrounding temperature. The enhancer DNA conformation was the only variable in these experiments and, therefore, most likely was responsible for the observed difference in the tWt/tSt values in the 10- and 40-min reactions. A wild-type enhancer that was moderately curved just prior to the start of transcription more positively influenced transcription of the E1A gene than the same enhancer when it was highly curved. The curved DNA conformation in the E1A enhancer clearly affected the stimulation of the E1A gene transcription.

DISCUSSION

In vitro transcription assay systems which could distinguish conformational effects of DNA were established. Using these systems, it was shown that the curvature in the Ad2 enhancer is an architectural element which is required for transcription activation of the E1A gene. It was also revealed that an exogenous prokaryotic DNA curvature located more than 110 bp upstream from the TATA box can repress or enhance transcription of the E1A gene. After discussing several points with regards to in vitro assays, possible mechanistic roles of DNA curvature in enhancing transcription will be considered.

Enhancer effects are marked in vivo but not in vitro. Nonetheless, there has been some success in establishing in vitro enhancer activity. The SV40 enhancer can stimulate transcription in HeLa cell extracts, usually about 10-fold, when positioned relatively close to the initiation site (26, 27). The immunoglobulin heavy chain enhancer can preferentially stimulate transcription in B cell extracts (28, 29). The most marked effect of this enhancer so far was a 15-fold stimulation of transcription (28). Deletion of the E1A enhancer, even in vivo, resulted in only a 20-fold decrease in cytoplasmic E1A mRNAs (19). Considering the disparity between in vivo and in vitro enhancer effects, the 2-fold stimulation of transcription by the E1A enhancer may represent a near-maximal effect in vitro. The conformational effect of the E1A enhancer that is shown in Fig. 5 may be much greater in vivo.

The unexpected transcripts that were produced did not result from degradation of the 5-end region of the run-off transcripts, because faithful transcript from the tWt and that from tSt have the same 5-end structure (Fig. 5B). Another possibility can be raised that the degradation might have occurred in the 3-end regions. This is also unlikely because the test and the reference templates shown in Fig. 2 and the template tWt shown in Fig. 5 all produced truncated transcripts irrespective of the sequence difference in their 3-end regions. Thus, it seems that the unexpected transcripts were transcribed from an unexpected initiation site located only several bases downstream from the authentic initiation site of the E1A gene. Some transcription factor in the HeLa cell lysate, which is required for faithful transcription, may have been labile around or above room temperature. However, I cannot answer why only the template tSt did not produce such transcript. The answer may lie in the structure of the template and the reaction conditions used in the experiment.

The DNA curvature in the E1A enhancer may present a framework for protein-DNA and/or protein-protein interactions. Alternatively it may influence DNA unwinding. The application of the first speculation to the results shown in Fig. 5 is as follows. The optimal DNA conformation for the E1A enhancer is presumed to be at human body temperature around 37 °C. Thus, the enhancer activity obtained from the experiment illustrated in the right panel in Fig. 5A was expected to be higher than the activity from the counterpart experiment (the left panel). This was substantiated (Fig. 5, C and D). In each experiment, the enhancer conformation should have changed to its 20 °C conformation immediately or soon after addition of the pSt2 digest to the reactions. During this very short period of time, the enhancer DNA conformation which was formed at 37 °C may have assembled or interacted with some transcription factor(s) more efficiently. Furthermore, only a slight difference in the amount or in the species of incorporated factors may have resulted in the difference in the extent of transcriptional stimulation. The framework hypothesis can also explain the low enhancer activities observed at 5 and 15 °C (Fig. 4B). The highly curved enhancer might be unable to provide a framework for transcription complex assembly.

DNA bend could play a more direct role in transcription activation. The second speculation is based on the intrinsic property of DNA curvatures. A theoretical study reported that the process of base pair opening is greatly facilitated by DNA bending (30). In prokaryotes, upstream DNA curvatures affect promoter melting (14, 17, 18). Upon binding of enhancer-binding proteins, the shape of the enhancer might have changed, resulting in different effects on base pair opening around the enhancer, which might have affected the promoter. The conformation formed around 37 °C might have been more effective in the opening process compared with that conformation formed at the low temperature. This hypothesis can explain the results shown in Fig. 4B and Fig. 5. It can also explain the stimulation of transcription by the DNA curvature of  origin (Fig. 4C). However, in this case, the existence of some protein factor which binds to DNA curvature with high affinity must be assumed, because the  curvature does not contain a cis enhancer element such as elements I and II. Nuclear extracts from HeLa cells contain such protein factors (25). Binding of such protein factors to the curvature formed at 5 °C might have resulted in effective base pair opening.

The negative effect of the  curvature at 15 °C (Fig. 4C) may be explained in terms of ``steric hindrance.'' The TATA-binding protein induces DNA bending (31). The  curvature formed at 15 °C might have sterically interrupted the TATA-binding protein-induced promoter bend, which might have led to an insufficient assembly of the components required for basal transcription apparatus and resulted in a reduced transcription level. Whatever the mechanism by which curved DNA conformation in the Ad2 E1A enhancer stimulates transcription, it seems safe to conclude that this curvature is a new member of curves with a function (32).