Modeling of DNA local parameters predicts encrypted architectural motifs in Xenopus laevis ribosomal gene promoter

Abstract

We have modeled local DNA sequence parameters to search for DNA architectural motifs involved in transcription regulation and promotion within the Xenopus laevis ribosomal gene promoter and the intergenic spacer (IGS) sequences. The IGS was found to be shaped into distinct topological domains. First, intrinsic bends split the IGS into domains of common but different helical features. Local parameters at inter-domain junctions exhibit a high variability with respect to intrinsic curvature, bendability and thermal stability. Secondly, the repeated sequence blocks of the IGS exhibit right-handed supercoiled structures which could be related to their enhancer properties. Thirdly, the gene promoter presents both inherent curvature and minor groove narrowing which may be viewed as motifs of a structural code for protein recognition and binding. Such pre-existing deformations could simply be remodeled during the binding of the transcription complex. Alternatively, these deformations could pre-shape the promoter in such a way that further remodeling is facilitated. Mutations shown to abolish promoter curvature as well as intrinsic minor groove narrowing, in a variant which maintained full transcriptional activity, bring circumstantial evidence for structurally-preorganized motifs in relation to transcription regulation and promotion. Using well documented X.laevis rDNA regulatory sequences we showed that computer modeling may be of invaluable assistance in assessing encrypted architectural motifs. The evidence of these DNA topological motifs with respect to the concept of structural code is discussed.

INTRODUCTION

The eukaryotic ribosomal RNA genes (rRNA) are commonly found in tandem chromosomal array, the so-called ribosomal DNA (rDNA), in which the transcribed regions are separated by intergenic spacers (IGS). Most sequences involved in rDNA transcriptional regulation are found within the IGS (the best-studied examples of IGS organization are reviewed in refs 1 and 2). In Xenopus laevis, reading in a 3′→5′ sense upstream of the start site, the following elements are found (Fig. 1).

Figure 1.

Structure of the X.laevis intergenic sequence (typical organization). Regulatory elements within the so-called non-transcribed spacer (NTS) that are indicated are the gene promoter (P), spacer promoters (sP), enhancer elements (e) made of 60/81 bp repeats (60/60+21), ‘0’ and ‘I’ regions made of 34 and 100 bp repeats, respectively. T3 is the site of transcription termination (T1 and T2 are processing sites of the transcript). The small black boxes indicate a region of 42 bp which is homologous between the enhancer elements and the promoters and present in both the 60 and 81 bp elements. Positions of oligo dT and oligo dA stretches are indicated. The 40S transcript is indicated as well as the direction of transcription. Zones 1, 2 and 3 refer to the active gene. They are regions of different chromatin organization (70). Region 1 has no nucleosomal organization but is covered with RNA polymerase I (∼1 molecule every 100 bp). Region 2 contains irregular arrays of nucleosomes. Region 3 lacks a nucleosome organization but binding of xUBF to the enhancer and the promoter leads to the so-called ‘enhancesome’ organization.

1. The RNA polymerase I (RPOI) promoter which consists of two essential domains, a core promoter proximal to the initiation site and an upstream control element (UCE) (3). Although the sequence between these two domains has a much smaller effect on the level of transcription initiation, helix spacing of the UCE and core elements appears essential to promoter activity and mutants with insertions or deletions of about half a helical turn are transcriptionally inactive (4).

2. The unique termination sequence element (T3) located ∼60 bp upstream of the 5′ boundary of the gene promoter. Independently of simply preventing promoter occlusion, the T3 element stimulates initiation following a mechanism which depends on the precise position of the terminator relative to the promoter, on the location of both elements on the correct face of the helix and on polymerase hand over (5–7).

3. The promoter-proximal spacer consisting of two active promoters upstream of an array of 60/81 bp repeat elements. Both spacer promoters have almost identical sequence and are flanked on the 3′ side by the same elements; nevertheless, the promoter upstream from the two blocks of 60/81 bp elements is sufficient to maintain full transcription efficiency and sequences just upstream have been involved in promoter dominance (8). Otherwise, the directly repeated 60/81 (60/60+21) bp elements enhance pre-RNA transcription in the absence of the active spacer promoter indicating that the spacer promoter only amplifies the enhancing effect of an array of 60/81 bp repeats (8–10). These enhancers appear to be associated into chromatin (11,12) and transcription from the spacer promoter may open up the enhancer chromatin to allow ribosomal gene activator to bind (8).

4. The promoter-distal spacer contains two repeated blocks of 100 and 34 bp corresponding to the ‘I’ and ‘0’ regions, respectively (8,13–15). Together and separately these 0/I repeats can function as enhancers of RPOI transcription and exhibit all of the known properties of the 60/81 bp enhancers. Deletion of the sequence between the 0/I repeats and the first spacer promoter seems to reduce the pre-RNA transcription suggesting that even this flanking sequence might play a regulatory effect although in a way not yet understood (16).

Transcription by RNA polymerase I requires the assembly of a multiprotein complex including the TATA-binding protein (TBP) which associates to form SL1/Rib1 (17), a transcription factor specific for the ribosomal gene promoter. Binding of the SL1/Rib1 complex to the ribosomal gene promoter is assisted by an upstream binding factor (UBF) which independently forms a complex with the promoter probably providing a binding site for SL1. Hence, the formation of the RPOI specific complex begins with the binding of UBF.

UBF shows sequence similarity with high-mobility group (HMG) proteins (18), an extensive family of abundant non-histone chromosomal proteins. The HMG box is an ∼80 residue domain that mediates minor groove DNA binding (19). The repeated HMG boxes of xUBF interact with the core promoter over the initiation site and downstream sequences in a colinear manner leading to DNA supercoiling and the formation of a disk-like structure, the ‘enhanceosome’ (20). In spite of its essential role in transcriptional initiation, UBF recognition has been shown not to display sequence selectivity except for a preference for dA- and dT-rich DNA, although it does bind strongly to supercoiled DNA and especially to potential cruciform structures (21).

Three-dimensional conformation of DNA has been associated with important biological processes (22). Evidence for diverse conformation of DNA comes from very different experimental approaches which included crystallography, gel migration, chain cyclization kinetics or nucleosome positioning studies (23–25). The single crystal study has highlighted the importance of the flexibility of the base pairs and the base steps (26) and has led to the definition and nomenclature of DNA parameters (27). Although the relative contribution of these parameters to the overall conformation of DNA was not absolutely evident, models and programs have been developed to assess structural motifs related to function as suggested from several analyses (reviewed in 28). In a preceding paper, we used computer modeling of eukaryotic ribosomal promoters in a large taxonomic group to show that common DNA structural features are maintained in spite of sequence divergence, suggesting that such structures are fundamental for RPOI function (29). We and others (30,31) extended these approaches to study RNA polymerase II promoters in order to highlight DNA structures involved in transcription regulation and to allow the development of algorithms able to identify promoters in genomic sequences.

In the present paper, we looked for sequence-dependent topological features which would account for IGS and gene promoter activities. In this respect, curvature, ‘bendability’, enthalpy and minor groove profiles of IGS were investigated. In addition, modeling of mutants used to delineate the gene promoter (4) was performed to identify local DNA changes critical in promoter activity and transcription initiation.

MATERIALS AND METHODS

Xenopus laevis ribosomal nucleotide sequence is from EMBL accession no. Y00132

Curvature

Three-dimensional co-ordinates of the helical axis are calculated along the sequence, as previously described (29,32), using parameters of the wedge model for curved DNA (33). The intrinsic bending magnitude is expressed as the ENDS ratio, defined as the ratio of the contour length of a segment of the helical axis to the shortest distance between its ends. ENDS ratios were computed using a window of 120 nt, moving along the molecule in steps of 1 nt.

Anti-bent DNA (straight DNA)

Anti-bent DNA detection is, like curved DNA, based on 3-D trajectory modeling. However, instead of looking for DNA curvature using only the classical twist, tilt and roll parameters for each successive nucleotide pair, twist angles are varied proportionally to their usual value to cover a whole range of theoretical DNA helix pitch values. The DNA helix trajectory is then calculated for each pitch value.

Using a sliding window of 120 bp, a series of ENDS ratio values is generated for each pitch value and plotted in three dimensions against the assumed pitch value and the position along the sequence. In many cases of interest, peaks appear in these ‘anticurvature maps’ for a particular position along the DNA sequence and for a particular helix pitch as previously described (34).

DNA duplex stability

The thermodynamic libraries that characterize all 10 Watson–Crick nearest-neighbor interactions in DNA (35) provide an empirical basis for predicting the stability (ΔG) of any DNA duplex region from its primary sequence. The local energy required for the strand separation of a 120 bp segment of DNA was calculated. Each plotted value takes into account the contribution of the surrounding nucleotides. All calculations were carried out with 1 bp step and parameters corresponding to 1 M NaCl, 25°C and pH 7.

Groove size

Three-dimensional co-ordinates of phosphate groups from strand 5′–3′ and 3′–5′ are used to calculate the shortest distance across the minor groove between a phosphate group taken on one strand and neighboring phosphate groups on the other strand. The value of 2.9 Å is taken as the radius of a phosphate group.

Bendability

Bendability was calculated using trinucleotide parameters established according to DNase I digestion efficiency (36,37).

RESULTS

Intergenic sequence (IGS) topology

Intrinsic curvatures fit functional subdomains. Sequence-induced DNA curvature or intrinsically bent DNA refer to net deflections of the helix axis from straight linearity. When curved sequences are repeated in phase with the helix screw, the local deflections point toward the same direction so that DNA appears curved. In contrast, DNA with out of phase curvatures looks like linear DNA. Figure 2A presents curvature profiles of X.laevis IGS deduced from ENDS ratio mapping. According to their amplitudes three types of peaks were found (solid line). (i) Higher curvature was associated with the spacer promoters (sP1 and sP2) and the gene promoter (P) and centered at positions –2158, –1048 and –18. One additional fragment with ENDS ratio peak similar to the spacer and the gene promoters occurred at position –232, in the region of the termination site. (ii) Furthermore four lower bends (peaks I–IV) centered at nucleotides –3868, –3148, –2462 and –1048 were clearly visible. Peak II, occurring within an AT-rich stretch, was seen to split the promoter-distal IGS portion into two subdomains: a 5′ domain consisting of the ‘0’ region and a 3′ domain containing the ‘I’ region. (iii) Lastly, the repeated elements were examined for inherent bends. The 100 bp elements exhibited curved motifs. Although very low, their values were clearly over the basal level assessed at 1.011 ± 0.004 using a random sequence showing a 80% GC content, as the X.laevis rDNA spacer. In contrast, the 34 bp repeats did not show any intrinsically bent pattern, even using smaller window sizes (34 and 20 bp, data not shown). In the promoter-proximal enhancer domain, regularly curved elements were observed corresponding to the alternate 60/81 bp repeats.

Figure 2.

(A) Curvature map of X.laevis ribosomal intergenic sequence. Nucleotide sequence was analyzed by computer modeling to reveal sequence-directed curvature. The 3-D helical path of the molecule was calculated using the wedge model of Trifonov. ENDS ratios were computed at a 120 bp window size and a 1 bp step as described in Materials and Methods. Positions of nucleotides are relative to the start point of transcription. Domains with enhancer function are indicated (‘0’, ‘I’, ‘60/81’). Regions containing the ribosomal promoters (sP1, sP2 and P) are boxed as well as interdomain regions. Low amplitude peaks are indicated using roman numbers. Position of anti-bent DNA sequence accessed as described in Materials and Methods is indicated. (B) Variation of duplex stability along the IGS sequence. The ΔG is calculated as the sum of nearest-neighbors interaction values for a 120 bp window sliding along the sequence. (C) Bendability variability along X.laevis IGS sequence. Values are calculated with DNase I based trinucleotide parameters (32) and computed at a 30 bp window size sliding at a 1 bp step.

Although the basic rules that govern intrinsic DNA bending are well established, our knowledge of the biological significance of these unusual structures is still incomplete. Using curvature modeling, we showed that IGS exhibits three kinds of intrinsically bent sequences. First, sharp curvature within the spacer promoters, gene promoter and termination site. Second, junction sequences exhibit moderate curvature that delineates structurally distinct subdomains. Third, the 100 and 60/81 bp repeated elements showed low but significant curvature patterns which might be related to their enhancer function.

Low helix stability in junction regions. DNA helix stability is the free energy difference (ΔG) between the duplex and the single-stranded states, and is governed by the nucleotide sequence. Free energy required for strand separation can be reliably calculated using experimentally determined thermodynamic parameters for the component nearest-neighbor dinucleotides (35).

Helical stability profiles for IGS sequence were modeled and the free energy required to strand-separate a given 120 bp DNA sequence is presented in Figure 2B. The minimum helical stability was shown to occur within the junction sequences between the ‘0’ and the ‘I’ regions, on one hand, and the ‘I’ region and the first promoter spacer, sP1, on the other. In contrast, high helix stability (342 kcal/mol) was reached at position –155 (at a GC only region), upstream of the 5′ end of the gene promoter and downstream of the termination site as compared to the basal level (317 ± 10 kcal/mol) deduced from a random sequence used as control.

Distribution of bending propensity within the IGS sequence. The ability to bend is thought to play an important role in different genome functions including DNA transcription and DNA replication. We used trinucleotide bending propensity parameters deduced originally from DNase I digestion data to analyze the variation in DNA bendability along the IGS sequence. DNase I, an enzyme with no pronounced sequence specificity, bends DNA towards the major groove. Since the cleavage rate appears to primarily depend on the flexibility the trinucleotide parameters are considered as indicators of bendability (37).

IGS sequence was seen to present a remarkable pattern of bendability parameters (Fig. 2C). Noticeably, a reproducible pattern of bendable elements, alternating from moderately bendable (∼5) to highly bendable (∼7), was found within the enhancer domains. In contrast, junction regions were found to contain more rigid and even extremely rigid sequences (∼2.5) as we can observe between positions –2451 and –2413.

Since DNA flexibility assesses the ability of DNA to respond to stimuli by changing its conformation and/or topology (28), these strong variations shown at privileged positions within the IGS may be of great importance in gene regulating function.

Winding trajectory and right-handed chirality of enhancers. In order to assess the spatial organization of the repeated elements, we modeled the 3-D trajectory of 34 bp, 100 bp and 60/81 bp decamers. Figure 3 presents the mean-plane trajectory of each of three different repeated element regions showing a slightly winding pattern reminiscent of superhelicity in 3-D space. Consistently, an orthogonal view of the same pictures reveals higher order structures with a right-handed chirality for all three repeats. In all these cases one complete turn needs about 5 units. Although the number of the repeats may greatly vary according to the individual, it is always more than five as in the typical IGS structure shown in Figure 1.

Figure 3.

Computer-generated molecular models of tandem arrays based on 34 bp, 100 bp or 60/81 bp repeats which are repeated motives from ‘0’, ‘I’ and ‘60/81’ enhancer domains. Five repeat units of each of these domains are shown in (A), (B) and (C) respectively (although 10 units were modeled, the number is limited to five to avoid superimposing the spires in the figure). The left part of the figure is a longitudinal view of the nucleotide sequence trajectory and the right part corresponds to an axial view. The axial view clearly shows the superhelical organization of these oligomers. All these superhelices were seen to be right-handed. (C′) Transition from the right-handed inherent superhelix to a left-handed induced superhelical form. The figure shows the calculated shape of the 60/81 bp repeat under gradual increase of superhelical density.

In order to better understand the function of the 60/81 bp enhancing sequence, we have analyzed the 3-D shape of different constructions which were previously demonstrated to maintain or lose their activity (38,39). The artificial enhancers that were constructed by polymerizing shorter or longer elements but keeping intact the central part of the natural repeat maintained their activity. Although the nucleotide sequence was changed we have seen that all maintained a right-handed superhelical shape, which came by repeating the same curved element. In contrast, mutating the central region by changing only 5 nt abolished the function (5′-CGGGC-3′ was replaced by 5′-TCTAG-3′) (38,39). In this case we found that the curvature was strongly altered, thus suggesting that the curved element plays an important role in the enhancing activity.

Because transcription of the ribosomal gene is particularly active we studied the effect of negative supercoils produced in the wake of polymerase progression. Figure 3C′ shows the effect of untwisting on the 3-D organization of the 60/81 bp enhancing sequence. A remarkable transition is observed when the superhelical density (σ) reaches –0.07 resulting in a planar conformation. Further untwisting is accompanied by the change of the right- to a left-handed supercoiled organization. The same transition for the same superhelix density is observed in each of the two arrays of the 60/81 bp enhancing sequence from the wild-type IGS. A lower value (σ = –0.04) is obtained for the 100 bp enhancer. These results, and our study of the artificial enhancers tested by Pikaard and Reeder (38,39), indicate that the degree of supercoiling density necessary to obtain the transition form is a property of the sequence. For instance, the so-called ‘33 bp’, which is made by the polymerization of the core promoter sequence, is seen to change its chirality for a σ value of –0.03, while the ‘26 bp’ artificial enhancer, which consists of only the central region of a natural 60 bp enhancer, needs a superhelical density of –0.17. Finally, this effect is not observed with constructs deprived of curvature as is the case for the mutated structures which were obtained by alteration of the central region.

Architectural signatures in core promoter and xUBF recognition

Bent alteration and minor groove profile of gene promoter of transcriptionally active and inactive variants. Bending toward the DNA minor groove diminishes groove width while the major groove opens up (40). We compared minor groove size profiles in relation with intrinsic bending of the wild-type promoter with those of linker scanning (LS) variants (LS –36/–28, LS –29/–22, LS –18/–9, LS –8/+1, LS +7/+10) (4). As shown in Figure 4, minor groove narrowing, extending from nt –20 to –29, was observed within the control core promoter with narrowest width at positions –24 and –25. Base exchanges in mutants LS –36/–28, LS –18/–9, LS –8/+1 and LS +7/+10 had no significant impact on the general minor groove profile, but LS –29/–22 showed a significantly expanded minor groove around positions –24 and –25 (Fig. 4B and D). As these unusual properties concern an oligo(dT)·oligo(dA) stretch (dT₆·dA₆), our data are consistent with the experimental approaches used to analyze A-tract conformation and showing that (i) four (or more) consecutive adenines do display minor groove narrowing (41) and (ii) the introduction of a single G, C or T within an A-tract is sufficient to abolish this effect and simultaneously to reduce curvature (42).

Figure 4.

(A and C) Two-dimensional projections of the spatial path of the promoter region from wild-type and –36/–28, –29/–22, –18/–9, –8/–1, +7/+10 mutants nucleotide sequences. The mean bending angle (α) calculated from the 3-D co-ordinates of the trajectory of every sequence is indicated. The core promoter is comprised between the two grey spots which encompasses the black spot (start point of transcription). UCE nucleotide sequence is delimited by the two white spots. (B and D) The size variation of the minor groove width. This size is calculated from the 3-D co-ordinates of phosphates groups computed using the wedge model from Trifonov. Changes in the sequence are compared to the wild-type.

The role of intrinsic curvature in promoter activity has already been intensively studied and different roles that it may play in promoter function have been suggested (43). We compared the wild-type promoter to the cognate LS variants (LS –36/–28, LS –29/–22, LS –18/–9, LS –8/+1, LS +7/+10) used to delineate the core region in surrogate genetics (4). All these variants are inactive but one was shown to sustain an elevated rate of transcription. The control promoter shows a strong curvature centered at position –26. Curvature angle values calculated from residue –85 to residue +35 are presented in Table 1 in relation with transcription efficiency (4). In this respect, variants LS –36/–28, LS –18/–9, LS –8/+1 and LS +7/+10 which did not display any promoter activity, exhibited a similar curvature to the wild-type. In contrast, LS –29/–22 whose transcription rate was essentially wild-type control (0.9 times) has a more flattened curvature profile (Fig. 4A and C).

Table 1. Intrinsic angle of curvature compared to transcription efficiency in wild-type promoter and LS variants of the core region.

	Wild-type	LS –36/–28	LS –29/–22	LS –18/–9	LS –8/+1	LS +7/+10
Intrinsic bent angle value	54°	49°	28°	53°	48°	49°
Transcription efficiency	1	<0.01	0.9	<0.01	<0.01	<0.01

Transcription efficiency values for LS variants are from Mougey et al. (16).

The result showing that transcription may be abolished although the promoter curvature remains unaffected does not conflict with previous results on the role of curved elements in promoter function since the promoter function may be altered in many different ways. But the surprise comes from the LS –29/–22 mutant. Previous results have shown that the core promoter between –16 and –33 must comprise an extensive functional region, since every point mutation resulted in a loss of promoter activity (44). This loss of activity is well explained by knowing that the region from –5 to –30 is a binding site for transcription factor(s) (45). Strikingly, our data show that efficient transcription may occur whether or not the template displays curvature, whether or not the binding site is altered. In order to tentatively explain the result, it may be hypothesized that sequence-induced structural changes in LS –29/–22 could mimic putative promoter reshaping resulting from pre-initiation complex binding in the wild-type. Base changes in the LS –29/–22 variant would induce favorable conformational changes (e.g. curvature, groove size, exposed groups etc.) so that LS –29/–22 would behave as a ‘permanent’ promoter.

Spacing of minor groove narrowings on either side of the initiation site. xUBF is believed to be the initial step in the formation of the RPOI holoenzyme, despite its lack of DNA sequence specificity. xUBF binding however occurs in a phased manner within the core promoter and continues well downstream of the initiation site (3,20,46). Like other HMG proteins, xUBF interacts with the minor groove of DNA (47) and HMG box 2 is known to bind over a (dA)₆ tract downstream and a (dT)₆ tract upstream of the initiation site (2). Boxes 1–3 are required for a full DNA looping and this corresponds with the minimal protein region shown to have activity in in vitro transcription (48).

In order to search for pre-shaped structure of the core promoter that might account for xUBF positioning, we modeled the minor groove profile of the gene promoter and downstream sequences up to position +60 (Fig. 5). In addition to narrowing at position –24/–25 as reported above, the width of the minor groove was modified at position +38/+39 (second half of box 2 site) in the transcribed region. These motifs fit the non-overlapping sequences between –35 and +10 on the one hand and +18 on the other, found to correctly position xUBF across the 40S initiation site (31). Accordingly, these positions closely flank the minimal regions found to be sufficient for phase xUBF binding.

Figure 5.

Size variation of minor groove width calculated along the nucleotide sequence. Start position is –60 from the start site of transcription and end position is +60 from this site. Position of xUBF binding boxes are indicated.

Thus, as suggested by the transcriptionally active LS –29/–22 variant, RPOI holoenzyme may require promoter reshaping to initiate transcription. It can be hypothesized that xUBF binding reverses the intrinsic bend of the gene promoter within its HMG box 2 binding site which is determined by the –24/–25 and +38/+39 minor groove narrowing signatures as was reported for HMGI(Y).

DISCUSSION

Genome packaging and highly efficient regulation of gene expression are both accomplished through a very complex supramolecular organization of the DNA. Inherent sequence dependent characteristics such as curvature, flexibility, helix stability or local variation in groove shapes together with modifications induced by proteins determine the final organization of regulatory DNA segments. The results presented here are an attempt at analyzing the initial shape and underlying properties of a gene regulatory region, the one involved in RNA polymerase I transcription.

Using computer modeling of the rDNA gene promoter and upstream regulatory DNA sequences, we showed that intrinsic bends, spread over the IGS elements, delineate subregions with distinct structural properties and could account for their differential functional behavior. The two spacer promoters of the X.laevis IGS as well as the gene promoter display equivalent intrinsic curvature despite their very distinct functional roles. Upstream flanking sequences have been invoked to explain differences between the two spacer promoters (8). They do exhibit very different structures. In reviewing data on non-B structure’s characterization and their role in transcription regulation, Van Holde and Zlatanova (49) have suggested that these structures could constitute a transcriptional block that could be removed through protein binding (50,51). Within the IGS, non-B structures could be involved in the interaction of the far-upstream enhancers with their cognate promoters and the formation of stable initiation complexes at the promoter (see 1 for a recent review). IGS junction elements could be responsible for fine tuning these promoters, through a scenario reminiscent of the concept of ‘conformational compensation’ (49), which accounts for the role of non-B structures in accommodating the overall superhelicity in a dynamic equilibrium between active and inactive promoter through protein binding. Our results may thus explain the dramatic reduction of pre-RNA transcription upon deletion of the 3′ joining sequence located just upstream of the first spacer promoter (16).

Our data indicating a right-handed superhelical organization of the enhancing sequences are consistent with a role of sequence-induced superhelicity in enhancer function (52). As the passage of a transcription complex generates the accumulation of supercoils, relaxation of this torsional stress may be essential to maintain the ribosomal transcription process at its full rate. It is then tempting to speculate that the intrinsic supercoiled organization of enhancers co-acts with topoisomerase I to accommodate transcription with the torsional tension imposed on DNA (53). Their joined function might give a decisive advantage that will render possible the extraordinary levels of transcriptional activity which is a characteristic of the ribosomal gene.

We report here that enhancing sequences from X.laevis are highly flexible structures and possess an unusual right-handed sequence-induced organization. This feature is not unique to this species. Like X.laevis, enhancers of RNA polymerase I in higher eukaryotes are all made of repetitive elements within the IGS of the rRNA genes and we found them to have the same right-handed supercoiled organization (data not shown). Moreover, we have also verified that the different constructions exerting enhancing function in the rRNA gene from X.laevis that were previously tested (38,39) also possess the same right-handed supercoiled intrinsic organization.

Another report (54) has demonstrated that the 60/81 bp enhancer from X.laevis can be exchanged with a plant enhancer (Arabidopsis thaliana) without changing the function. Thus we suggest that information from X.laevis may be also valuable to better understand the ribosomal enhancing function in other higher eukaryotes.

The analysis of the influence of untwisting of natural and artificial (functional and non-functional) enhancing sequences on the conformational response has afforded the following important information: upon untwisting, the right-handed superhelical DNA gradually increases its radius of curvature until becoming planar. Further untwisting converts the molecule to a left-handed superhelix, which in turn progressively decreases its radius of curvature. This is strongly dependent on the initial sequence-dependent trajectory: upon untwisting, the intrinsically right-handed supercoiled DNA may change its handedness whereas a left-handed superhelix will become more and more supertwisted with smaller radii. Very similar modifications have been described (55) for ethidium bromide untwisting of intrinsically supercoiled synthetic DNAs.

Taking these observations into account, in contrast to the right-handed structure, an inherently left-handed superhelix is not expected to support the high rate of transcription of the ribosomal gene. Because of the rapid accumulation of supercoils on already supertwisted DNA, a left-handed organization upstream of the polymerase could rapidly hamper polymerase activity. The RNA polymerase I progression should then entirely depend on the function of the topoisomerase which then would need to follow the extremely high rate imposed by the function of the rRNA gene.

Different properties may be associated with this inherent supercoiling that may act at different levels to facilitate ribosomal gene transcription. (i) Topoisomerase I studies have shown that two different sites, called S and C, may be used by this enzyme to cleave DNA. The S sites are induced by DNA supercoiling. In contrast, the C sites are constitutive and were seen to function on relaxed and linear DNAs. These last sites were found to be specifically localized within intrinsically bent DNAs (56,57). Thus, inherent supercoiling may form permanent sites of cleavage for topoisomerase I. Moreover, upon induced supercoiling, domains that contain such structures are seen to be preferentially topoisomerized as a consequence of a higher rate of association of the enzyme with intrinsically bent DNA sequences (56). (ii) Accumulation of negative supercoils in the wake of the RNA polymerase may cause a progressive unwinding of the 60/81 bp region. However, we can see in Figure 3C′ that right-handed inherent supercoils adopt a progressively more planar conformation, and at a determined superhelical density, adopt a left-handed conformation. In contrast to a left-handed organization, the right-handed inherent supercoiling of the enhancer region may, upon DNA untwisting, increase their radii of curvature. This change may proceed until a shape transition is reached which is immediately followed by the acquisition of a left-handed organization. This type of response of DNA to untwisting may increase the limit above which accumulated tension is expected to block further transcription. Such behavior might allow RNA polymerase I to work at full load, which is a characteristic of ribosomal gene transcription. (iii) The natural superhelical conformation of the DNA could also enhance the affinity of proteins for DNA binding and hence stimulate transcription (38,58,59). Consistent with this latter possibility is the case of the transcription factor xUBF, which binds the upstream element of the promoter. This regulatory protein, which seems to recognize DNA conformation rather than a particular sequence, shows a strong binding preference for supercoiled DNA (21) and has multiple fixation sites within the 60/81 bp enhancing sequence (60). However, it must be noted that xUBF binding alone is not sufficient to explain the function since the mutated enhancers that were shown to lose function retained the ability to bind xUBF (39). It is nevertheless most likely that in the ribosomal gene the precise mechanism of action of the enhancer depends on the interplay of reciprocal modifications of conformation of both DNA and associated proteins.

Moreover we found that, depending on the structure coded by the sequence, the transition from the right- to the left-handed organization may occur with different superhelical density indicating that the precise function does not only depend on the right-handed chirality.

Thus, the inherent sequence-dependent flexibility (bendability and twistability) and characteristics of inherent supercoiling (chirality and reactivity to superhelical density) together with deformations induced by proteins (xUBF, and possibly other proteins) might determine the final shape and regulatory properties of the region.

Further modeling of the gene promoter allowed us to point out minor groove features that could be involved in transcription control and to speculate that minor groove narrowing on either side of the 40S initiation site could constitute a recognition element for correct positioning of xUBF. As HMG box 3 was necessary for functional adjustment (48), we looked for motifs in the HMG box 1–3 construct that could explain such a requirement. We found a putative AT-hook element (61), located at amino acids 285–287 within the linker connecting HMG box 2 to HMG box 3, which could facilitate HMG box 2 to fit deeply in the narrow minor groove. Otherwise, HMG box 3 sequence could be directly involved in this synergistic positioning. Crystallographic analyses of HMG protein–DNA complexes have revealed minor groove widening through partial insertion of one or more hydrophobic side chains of HMG domain into the base stack to pry open the minor groove and kink the DNA molecule (62,63). According to our data showing constitutive minor groove opening and helix distortion of the transcriptionally active LS –29/–22 variant promoter, we hypothesized that similar mechanisms would be involved in rDNA transcription initiation through xUBF binding to the wild-type gene promoter. As reported, xUBF might reverse the rRNA gene promoter bend, notably through HMG box 2 recognition and binding to the –24/–25 and +38/+39 minor groove narrowing signatures; as a consequence, the targeted minor grooves should be extensively expanded.

Inherent sequence dependent deformability, as well as deformations induced by protein binding, both determine the final DNA 3-D structure involved in chromatin function (64). The existence of a structural code embedded in the genetic code was first proposed by Calladine (65) who stressed the possible importance of topological motifs delineated by base pair bending and shearing along and across the helix axis, suggesting an architectural or 3-D level superimposed to the linear genome sequence. In this respect, changes in intrinsic curvature of promoters by transcription factors and architectural proteins have been shown to be a basis for transcription regulation (63). This also accounts for the remarkable specificity of some pleiotropic transcription factors such as NF-κB, ATF-2/c-jun and p53 (66–68). Accordingly, sequence-dependent features add a new perspective to the ‘literal’ analyses of genomic sequences (69) and could be useful in detecting elements such as promoters (29) as well as other regulatory sequences.

Our approach thus offers a new perspective to search for encrypted motifs in DNA conformation and possibly reveal a higher-order code involved in the dynamic organization of the genome.