ABSTRACT
Nucleotide sequences of DNA regions containing eukaryotic ribosomal promoters
were analysed using strategies designed to reveal sequence-directed structural features. DNA curvature, duplex stability and pattern of twist
angle variation were studied by computer modelling. Although ribosomal
promoters are known to lack sequence homology (unless very closely related
species are considered), investigation of these structural characteristics
uncovered striking homologies in all the taxonomic groups examined so far. This
wide conservation of DNA structures, while DNA sequence is not conserved,
suggests that the determined structures are fundamental for ribosomal promoter
function. Moreover, this result agrees well with the recent observations
showing that RNA polymerase I transcription factors have not evolved as
intensively as previously suspected.
Unlike RNA polymerases II and III, RNA polymerase I is involved in the synthesis
of a sole product, pre-ribosomal RNA. Consequently, it requires recognition of only one kind of
starting signal for the expression of hundreds of gene units. It is highly regulated to be responsive to both general metabolism (e.g. growth rate) and to specific environmental challenges (see
1
-
3
for reviews). Surprisingly, systematic analyses of the nucleotide sequence around the origins
of transcription of rDNA in different organisms has revealed no common pattern
of nucleotide sequences (
4
-
6
). Moreover, the RNA polymerase I transcription system appears to diverge
considerably between organisms. Ribosomal transcription is generally specific
to taxonomic orders, the promoter of one group not being recognized by the transcription factors of another. This disparity of RNA polymerase I promoter sequences is apparently in agreement with this order of species
specificity (
6
,
7
).
However, several lines of evidence now suggest the existence of a common
organization of all the promoters (
7
-
25
). According to these results, a ribosomal promoter consists of essentially two
domains. There is a `proximal promoter domain' (also called the minimal or core
promoter) of ~45 bp, which includes the start point of transcription and is absolutely
required for determining the accuracy of initiation, and an `upstream promoter
domain' or upstream control element (UCE), mapping at about -150 bp relative to the transcription start site.
Moreover, recent findings indicate that the RNA polymerase I transcription
system has not diverged as intensively as first appeared. For example, one
ribosomal transcription factor, the upstream binding factor (UBF) was found in
human and also in
Xenopus
, mouse and rat (
26
-
30
). Mouse UBF and human UBF (and other transcription factors) were found to be
functional on either the mouse or human promoter (
28
). A simple change of half a helical turn is also able to convert a
Xenopus laevis
promoter into a highly active mouse promoter (
18
). This example implies that proteins as well as DNA from both species share
homologies, despite the divergence between the rRNA promoters. These apparently
conflicting results can be easily explained if, as recently proposed for
polymerase II transcription (
31
), the spatial organization of ribosomal gene promoters plays an important role
in species specificity.
General studies on transcription have focused on protein-protein interactions as playing a critical role both in promoter recognition and
in regulation of transcription. These studies have also pointed, as indicated
before, to an additional mechanism, the assembly of a stereospecific
nucleoprotein complex. This process requires proteins that bind to DNA in a
sequence-specific manner, but function as architectural components. Thus, a
different spatial organization of modular elements might induce species specificity.
Initiation involves sequence-specific binding of transcriptional factors to DNA. Stereospecific
assembly of the nucleoprotein complex requires, in addition, that DNA
structures facilitate, or at least allow, the architectural complex to be
built. Because they are likely to represent a physical support for promoter
activation, it is of major interest to test the hypothesis that specific
structural features might be present within a promoter domain and conserved
throughout all taxonomic groups.
The number of works focusing on the potential role of DNA structure in the
maintenance of a specific function is now increasing. It has been shown that
sequence-directed bending of DNA causes local variations in the structure of
genomes (
32
). Bent helices are characteristics of some promoters and of other regulatory
regions (for reviews see
33
,
34
). Moreover, the basic rules of DNA curvature are now well enough established to
render this parameter directly accessible to analysis on the basis of the DNA
sequence (
32
,
35
,
36
and references therein). Direct examination of the nucleotide sequence has also
proven to be valuable for the study of other structural parameters of the
helix, such as duplex unwinding elements (
37
,
38
), variations of twist angle (
39
,
40
) and variations of groove size (
41
).
In this context, we have chosen to analyse these widely studied structural
parameters (DNA curvature, helical stability and unusual variations in twist
angle values) instead of directly comparing the various DNA sequences, as has
been done in the past. This allowed us, as a contribution to the understanding
of the role of DNA structure in the function of eukaryotic ribosomal promoters, to realize a comparative study of sequence-directed structural features and to examine how they can reflect specific
structural properties.
Our results confirm a basic conserved organization of ribosomal promoters into domains. We show that these domains may be distinguished on the
basis of their DNA structural features. These results support the existence of
a modular organization of the ribosomal gene transcription apparatus and underline the importance of the spatial organization of the underlying DNA. Because of the conservation
of these structural features from lower plants to human while the DNA sequence is not conserved, the results support the view of a
structural code for DNA regulation sequences. This code should correspond to
DNA structures necessary to provide a physical support for the transcription
machinery.
Nucleotide sequences are from the GenBank/EMBL database:
Homo sapiens
(X01547),
Rattus norvegicus
(X00677, K01588, M12030),
Xenopus laevis
(J01005),
Xenopus borealis
(X05263, Y00132, X00184),
Drosophila melanogaster
(X02210),
Paracentrotus lividus
(X63234),
Tetrahymena pyriformis
(J01212, M10096),
Dictyostelium discoideum
(X00601),
Arabidopsis thaliana
(X15550),
Pisum sativum
(X52575),
Triticum aestivum
(X07841),
Zea mays
(X03990)
and
Physarum polycephalum
(
42
) The program Geneworks was used to find the best sequence alignments and to
calculate the percentage homology between the DNA molecules analysed here.
The algorithm for calculating DNA bending from nucleotide sequences was
published by Eckdahl and Anderson (
43
). Three-dimensional co-ordinates of the helical axis are calculated along the sequence as
previously described (
44
) using the parameters of the wedge model for bent DNA from Ulanovsky and
Trifonov (
45
), Bolshoy
et al
. (
46
) and de Santis
et al
. (
47
). The magnitude of DNA bending on curvature maps is expressed as the ENDS
ratio, which is 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 at a window width of 200 nt and with a window step of 10 nt to allow
comparison of the results with the data of Anderson and co-workers (
32
). High resolution analysis within curved regions was performed with a window
width of 30 nt and a 1 nt step. This window size was chosen to be large
compared to the helix pitch so that very local variations are not taken into
account but remains small in comparison with promoter size, which is only ~150 nt, and far less (~45 nt) for the core promoter.
The thermodynamic library of Breslauer
et al
. (
48
) characterizing all 10 Watson-Crick nearest-neighbour interactions in DNA was used to calculate DNA duplex
stability. These thermodynamic data provide an experimental basis for
predicting the stability ([Delta]
G
) of any DNA duplex region by inspection of its primary sequence. We have
developed a computer program similar to the Thermodyn program of Kowalski (
38
) to calculate the mean sliding [Delta]
G
for the chosen size of DNA segment to be studied. Each calculated value takes
into account the contribution of the surrounding nucleotides. Here, values
refer to the disruption of the interaction in an existing duplex at 1 M NaCl, 25oC and pH 7.
Variations of the twist angle were mapped as described for the calculation of
duplex stability. Twist angle values were taken from Kabsch
et al
. (
39
) and from de Santis
et al
. (
47
).
Calculations were made with the PACS DNA program developed in our laboratory and
already exploited in various nucleic acids studies (
44
,
49
,
50
,
51
).
An increasing number of works are focusing on the presence of intrinsically
curved DNA in regulatory regions (
33
). In order to test whether bent DNA is also an important structure of the
ribosomal promoter we have analysed the structure of the corresponding DNA fragments by computer modelling. Independent wedge models of DNA curvature, like those of Trifonov and of de Santis (
45
-
47
,
52
), were used in this study. These models were shown to be reliable for the
prediction of electrophoretic retardation and circularization and were also
used for theoretical prediction of nucleosome positioning (
36
,
43
,
44
,
47
,
50
,
51
,
53
-
59
).
Figure
2
shows an analysis of DNA curvature within the intergenic ribosomal spacer of
sequences from both the animal and vegetal kingdoms. When available these
nucleotide sequence analyses span from 3 kb upstream to 1 kb downstream of the
transcription initiation site. This allows not only fine analysis of the
ribosomal gene promoter, but also of a large region of the non-transcribed spacer (NTS) containing the spacer promoters. The magnitude of
bending is expressed as the ENDS ratio and was computed for a window size of
200 bp, thus allowing a comparison with the values reported by Van Wye
et al
. (
32
). This analysis allowed us to determine whether curved elements are frequent or
unusual features in the surroundings of the promoter. As a comparison, Van Wye
et al
., in their analysis of the GenBank/EMBL database, defined values above 1.5 as
strong bending elements (for example, the bent motif associated with the yeast
ARS1 has a value of 1.54). Figure
2
shows that most of the ribosomal gene promoters display a significant DNA
curvature. The deflection of the helix axis is notably stronger in some species
(about 1.7 for
P.sativum
, 1.6 in
D.discoideum
) than in others. This observation is consistent with the species-dependent pattern of bending described by Van Wye. Promoters with a high
G+C content have low bending scores, thus resembling on this point bacterial
G+C-rich ribosomal promoters (
32
).
Because 200 bp is large compared to the size of a promoter (the core promoter is
reported to be ~45 bp) we used a smaller window size (30 bp) to investigate more precisely
the organization of rRNA promoters. It should be noted that the ENDs ratio
values calculated here cannot be directly compared with previous ones (since
bending results from the cumulative contribution of small curvatures in phase
with the helix pitch, the ENDS ratio depends on the window size). Using a
smaller window size allows us to separately visualize these small curvatures
and enables us to see small stretches of structural elements otherwise
undetected.
Moreover, since many available DNA sequences are often limited to the nucleotide
sequence of the promoter, decreasing the window size allows the analysis of a
larger number of promoters. Figure
3
shows that several minor bending elements are involved in the three-dimensional shape of the promoter. A segment of non-curved DNA is also observed around the transcription start.
Strikingly, this straight motif is highly conserved in evolution, as indicated
by its low standard deviation in the averaged curves (B and D). This is clearly
visible whatever the model used for structural prediction. It is worthwhile
noting that it is essentially the structure and not the sequence that is
conserved around the initiation site. Although a 13 bp conserved region
surrounding the transcription origin was found among
Xenopus
species
X.laevis
,
X.borealis
and
X.clivii
(
61
), no significant homologies were detected in more distantly related organisms
(human, mouse,
Xenopus
,
Drosophila
and
Tetrahymena
) (
4
).
The thermodynamic stability of double-strand DNA molecules is sequence dependent. Not only GC% but also nearest-neighbour interactions between the DNA bases are involved in DNA
duplex stability. Breslauer
et al
. (
48
) have characterized all the 10 possible interactions in a Watson-Crick DNA duplex structure. They have also shown that the stability of
the duplex structure can be considered to be the sum of its nucleotide nearest-neighbour interactions. Their data are used here to predict the relative
stability of local domains within the DNA region containing the ribosomal gene
promoter. The approach is very similar to the one realized by Umek and Kowalski
(
37
,
38
,
62
) for characterizing duplex unwinding elements (DUE) in replication origins.
Although the overall [Delta]
G
value appears to vary widely from one species to another (
Xenopus
rDNA is G+C rich, but
Drosophila
and
Tetrahymena
have low G+C contents) it is noticeable (Fig.
4
) that all the studied sequences show a decrease in [Delta]
G
values (and G+C%) in the region of the promoter. The extent of this decrease
may vary somewhat, but these values are always below the average [Delta]
G
of the surrounding sequences. It is also remarkable that this region of low [Delta]
G
is often flanked by a downstream stable domain (Fig.
4
b). As shown in the lower graph of Figure
6
, where a smaller window size (30 bp) is used, the decrease in [Delta]
G
values is due to a sharp decrease occurring essentially within the core
promoter domain and within the UCE. It is worth mentioning that the
transcription initiation site is localized at a transition zone between minimal
and maximal [Delta]
G
values. This `barrier of [Delta]
G
' is apparently conserved throughout evolution, even in the absence of sequence
homology.
Sequence-dependent variations in conformational parameters such as helical twist
(and other helicoidal parameters) contribute to the overall three-dimensional shape of the DNA surface and presumably to the ability of DNA
binding proteins to recognize specific sequences (
35
). Recently, MacLeod (
40
) has provided evidence for a so-called pyrimidine sandwich element (PSE) which seems to play an important
role in the interaction of
trans
-acting factors with DNA control regions. This shows that sequence-dependent variation in the pattern of the twist may be an important
structural feature involved in specific DNA-protein recognition and may play an important function in transcription
control.
Kabsch
et al
. (
39
) have shown that an angle larger than average usually tends to be compensated
for by a smaller angle in the immediately following dinucleotide. Sequence-directed variation in the twist angle tends to prevent accumulation of
over- or undertwisting along a DNA molecule. As a consequence, the structure of
the B-DNA backbone typically shows a gentle zig-zag of plus or minus a few degrees. In order to detect local
anomalies that may reflect some unusual structure, we have averaged successive
twist angle values and followed their variations along the molecules. A 200 bp
window size was chosen to focus on variations of large amplitude.
However, since the new evaluation of twist angles by de Santis
et al
. (
47
) resulted in values largely different from those of Kabsch, we used both sets
of values and compared the two results. Figure
5
shows a comparison of this twist angle pattern in
X.laevis
,
X.borealis, P.lividus, D.melanogaster, D.discoideum, A.thaliana, P.sativum,
T.aestivum, Z.mays
and
P.polycephalum.
The choice of these species was dictated by the size of the available
sequences. Comparing long nucleotide sequences allows us to see to what extent
the observed patterns are different from those of neighbouring nucleotide
sequences. Although there is a lack of sequence homology and these sequences display an extremely
different G+C%, we can see in the figure that the profiles of twist variation,
characterized by a successive accumulation of over- and undertwisting, are strikingly similar irrespective of whether Kabsch or de Santis values of twist angles are used (Fig.
5
A and C). Moreover, in most cases a sharp and continuous decrease of twist value
is observed in the 200 nt sequence which includes the gene promoter. A possible
exception to this rule is
X.laevis
. However, it is possible to detect a similar event in Figure
6
(where a smaller window size is used), but the decrease is visible only in the
region of the core promoter.
Refining the analysis (30 bp window size) allows us to detail the structure of
the promoter. Figure
6
shows a structural map of the
X.laevis
ribosomal promoter compared to a similar analysis made on 12 promoters taken
from a wide range of taxonomic groups. The
X.laevis
promoter is omitted from this last analysis to avoid any interference with the
result. Since the
X.laevis
promoter is one of the most extensively studied (
16
-
20
), this allows us to position structural elements relative to the nucleotide
sequences important for promoter function. It is worthwhile mentioning that the
different functional domains of the
X.laevis
promoter (core, UCE, enhancer homologue) correspond to regions containing
specific structures. Although very different base compositions may be
encountered in nucleotide sequences,
X.laevis
is by far less A+T-rich (only 15%), comparison of the two types of graph reveals common
structural features. This reinforce the previous result that the sequences have
developed equivalent structural characteristics for assuming promoter function. This is particularly clear for the core promoter region, where we can observe exactly the same
pattern of variation (curvature, twist and helical stability) in
Xenopus
and in the set of 12 promoters. This result is indicative of a high structural
conservation of this promoter domain throughout evolution (while sequence is not, as previously shown in Fig.
1
). Although some species variation might be observed when individual patterns are considered, strong analogies are found throughout the promoter which allow us to distinguish different
regions. Reeder (
17
), using linker scanner mutagenesis, concluded that the
X.laevis
promoter is composed of three domains, one of which is an enhancer element. All
three domains are visible in our analysis and, although an enhancer element is
not described in other promoters than
Xenopus
, an intermediary structure possibly equivalent to it is observed in the set of 12 promoters.
The organization of structural elements within promoter domains supports the
proposal of Reeder (
17
) that the promoter functions as a set of interacting domains, but also suggests
that these domains are interdependent.
Transcriptional activation requires the ordered assembly of large multiprotein-DNA complexes. Important progress has been made in the identification and
purification of eukaryotic transcription factors, mainly dealing with RNA polymerase II. Analysis of the process of
initiation complex assembly has determined the complex nature of protein-DNA interactions and one remarkable outcome of this research is that DNA
not only contains information for binding cognate regulators but also has
intrinsic structural properties playing an active role in transcription
initiation (for a review see
31
). We report here an analysis of the intrinsic structural features of rRNA
promoters and discuss the possible functional involvement of DNA structure in
ribosomal gene transcription.
Several recent studies have shown that sequence-directed curvature and protein-mediated DNA bending play a key role in the regulation of gene
expression (for reviews see
33
-
35
,
63
,
64
). The first evidence of an intrinsic bending associated with the activity of a
bacterial promoter was obtained in 1984 (
65
) and the participation of bent DNA at nearly all the stages of prokaryotic
transcription is now well documented (
34
). Although rarely investigated in eukaryotes, intrinsically bent DNA has been
described in association with polymerase II promoters (
66
) and also detected in one polymerase I promoter, the ribosomal gene promoter
from
Physarum
(
60
). Here, we have found intrinsic bending as a constant component of the
ribosomal promoter. It is thus very likely that this DNA feature is a rule
rather than an exception in eukaryotic promoters.
The precise function of DNA curvature associated with ribosomal promoters has
now to be investigated, as was previously done with prokaryotic systems. Nevertheless, it is worthwhile noting that this
structural element may be involved in a large spectrum of functions. To give a
more precise idea of this variety it is important to note the following. (i)
Intrinsic bending may be involved in protein docking and/or in the wrapping of
DNA around proteins, which may thus be energetically favoured (see
67
for a review). (ii) It may also contribute to the formation of DNA-specific binding sites by modifying the groove shape, allowing the
exposure of residues that are to interact with the cognate protein (
41
, see
68
for a review). (iii) Bent elements are likely to affect chromatin structure
around the start site and should constitute a preferential binding site for HMG
box proteins like UBF (
69
-
73
). (iv) Bent DNA may also participate in local base pair opening when
destabilized by torsional stress or protein binding. In turn, once a base pair
is disrupted, unstacking creates a flexible joint which is very easily bent (
74
). (v) DNA intrinsic bending determines the three-dimensional helix path. The three-dimensional organization of DNA together with its flexibility are very important inherent properties that must be accommodated in the assembly of the stereonucleoprotein structure of the active
promoter (
31
, see
75
for a review). (vi) The relative orientation and phasing of bent elements may
be modified by variation in the superhelicity of the DNA, thus affecting the amplitude of the
curvature and, by the way, promoter activity (
76
).
Finally, it must be stressed that the same curvature is simultaneously involved
in several of these processes. Therefore, the assembly of an active promoter
must be viewed as a global and dynamic process involving the overall structure
of both proteins and DNA.
The process of initiation is dependent on localized melting of the DNA double
helix by the transcription complex. Several structural characteristics of the
DNA molecule
are known to facilitate base pair opening. Among them, DNA topology has been
shown to affect the thermal requirement for strand separation. DNA bending
described above is also known to lead to a significant decrease in the opening
energy. This is explained by a simultaneous lowering of the unstacking energy
and by the accumulation of energy within the sugar-phosphate backbone, which may be further released to open the DNA (
74
). In addition, DNA supercoiling and curved DNA associated with the promoter may
cooperate to induce a localized melting of the duplex (
35
,
77
,
78
). Kowalski and co-workers have recently shown that DNA unwinding elements are associated
with some prokaryotic promoters (the [beta]-lactamase gene) and that torsional stress alone is sufficient to locally unwind the DNA, even in the absence of
initiation proteins (
38
). The general [Delta]
G
decrease that we observed here within the promoter region of the rRNA genes locally lowers the energy
required for strand separation. It should be noticed that although A+T- and G+C-rich promoters do not have the same energy requirement for DNA
unwinding, they have the same necessity to open it in a well-defined region. Here we observed that [Delta]
G
profiles are very similar in A+T- and G+C-rich sequences and clearly indicate the position of the promoter
region.
The sharp [Delta]
G
increase 3' of the promoter is also an interesting point to discuss (although the
position and the amplitude of this peak may vary somewhat with the species). A
high [Delta]
G
might be related to either the specificity of initiation, through stabilization
of the initiation complex and/or to promoter clearance. It is worthwhile
observing that the region of high [Delta]
G
values overlaps the 3' region of straight DNA that surrounds the start of transcription. Base
pair opening in the same sequence of straight DNA needs far higher energy than
in bent DNA (
74
), so an increase in [Delta]
G
value and non-curved DNA might cooperate in positioning the region of double-strand opening. This might play a role in the accuracy of
initiation.
Twist angle variations in the ribosomal gene promoters analysed here are seen to
follow an unusual pattern when compared to surrounding sequences. This unusual
variation may lead to local variations in groove shape. Moreover, it should be
noted that the twist angle together with the other helical angles also
determine the three-dimensional helix path. Thus, this particular pattern of variation might
be associated with some basic three-dimensional organization that will be discussed later.
Earlier investigations to decipher the sequences necessary for transcription
initiation have revealed very little (if any) sequence homology between
species. Even related species (mammalian sequences) were shown to be no more
similar than two random segments with the same base composition (
4
). However, as stressed by Treco
et al.
(
4
), the rapid divergence of the sequences between species does not mean that they
do not serve a critical function, but only that a variety of nucleotide
sequences can carry out the same function equally well. The rapid evolution
might also be explained by fixation of selectively advantageous mutations. Our
results are in complete agreement with both explanations. Indeed, we observed
that promoter regions possess intrinsic structural properties which may play an
active role in the transcription process.
Previous reports have pointed out the importance of sequence variation in RNA
polymerase I promoters and underlined the fact that this variation implies a
concomitant evolution of the proteins of the initiation complex. However, these
proteins appeared to have evolved less intensively than expected (
28
,
79
). Therefore, our observation that a high level of sequence variation does not
necessarily imply important changes in the structure of the ribosomal promoter
(especially within the core promoter) is more in agreement with recent results showing that numerous transcription factors (UBF, TBP, TIF IA and TIF IC) are interchangeable between species. This
observation underscores the importance of transcription factors in determining
promoter selectivity (
79
) and again gives strength to the hypothesis that species specificity may arise
from stereoassembly of the nucleoprotein complex.
Because of its lack of sequence homology, the ribosomal promoter might be useful
in identifying the sequence-directed structural features that are fundamental for promoter function.
Here, we show that computer modelling analysis is a valid approach to identify
structurally active sites in rRNA promoters. These sites can be specifically
modified by mutagenesis, thus providing an additional experimental approach for
investigating the complex puzzle of transcriptional activation.
This work was supported by a research grant from the Association française pour la Recherche contre le Cancer (ARC). We thank A. Humbert for
helpful discussions and suggestions concerning DNA modelling. We gratefully
thank Dr S. Granjeaud and C. Blettry for their help in writing new programs.
REFERENCES
Return