ABSTRACT
The formation of a near complete loop of DNA is a striking property of the
architectural HMG-box factor xUBF. Here we show that DNA looping only requires a dimer of
Nbox13, a C-terminal truncation mutant of xUBF containing just HMG-boxes 1-3. This segment of xUBF corresponds to that minimally
required for activation of polymerase I transcription and is sufficient to generate the major characteristics of the footprint given by intact xUBF.
Stepwise reduction in the number of HMG-boxes to less than three significantly diminishes DNA bending and provides
an estimate of bend angle for each HMG-box. Together the data indicate that a 350
+-
16
o
loop in 142 +- 30 bp of DNA can be induced by binding of the six HMG-boxes in an Nbox13 dimer and that DNA looping is probably achieved by six
in-phase bends. The positioning of each HMG-box on the DNA does not predominantly involve DNA sequence
recognition and is thus an intrinsic property of xUBF.
Transcription of the ribosomal genes in eukaryotes uses a dedicated polymerase,
RNA polymerase I (pol I) and a dedicated set of transcription factors (for
recent reviews see
1
,
2
). Probably for this reason, the pol I factors and the promoters they recognize
display a high degree of species specificity. Despite this specificity,
promotion by pol I appears to be mechanistically similar in mammals and
amphibia. The promoter generally consists of two precisely spaced sequence
elements, the upstream control element (UCE) and the Core promoter element.
These elements have been well-defined by surrogate genetics in several organisms. They do not, however,
display significant sequence homology. The major steps in the assembly of a pol I pre-initiation complex are known (
3
-
8
). Recognition of the UCE and the Core promoter elements by the so-called upstream binding factor (UBF) allows the entry of a pol I-specific TBP-TAF complex, variously called SL1, TIF-IB, TFID, Factor D or Rib1. This leads to the formation
of a stable pre-initiation complex which can be specifically recognized by an activated
form of pol I. UBF, therefore, plays a key role in promoter recognition and in
pre-initiation complex assembly. Surprisingly, UBF displays little detectable sequence selectivity (
9
,
10
). Despite this, UBFs from human, mouse, rat and
Xenopus
recognize pol I promoters in a very similar manner.
Xenopus
UBF (xUBF) does not, however, support the entry of SL1 to the mammalian
promoters and the converse also appears to be true (
11
-
15
).
The UBFs contain multiple HMG-box DNA binding domains. These domains are a characteristic of a large
subgroup of architectural transcription factors (
16
,
17
). HMG-boxes are especially known for their capacity to induce severe kinks in
their DNA target sequences (
18
-
20
). XUBF contains five tandemly arranged homologies to the HMG-box domain (
21
,
22
) (Fig.
1
A). We have previously shown that xUBF binds to the transcription initiation
site of the
Xenopus
pol I promoter such that its HMG-box 1 protects bases -21 to -2 and +2 to +21 and HMG-boxes 2 and 3 protect bases downstream of +22 (
23
). Using electron spectroscopic imaging (ESI) we have directly measured the
stochiometry of single xUBF-DNA complexes, showing them to each contain two xUBF molecules (
24
). Most strikingly, the binding of a single xUBF dimer to the
Xenopus
ribosomal enhancers induces an Enhancesome complex in which a short segment of
DNA is looped in to a near complete turn (
24
). Together, the footprinting and ESI data strongly suggest that within the Enhancesome two xUBF molecules position themselves head to head along the DNA (Fig.
1
A). The manner in which xUBF induces DNA looping and the role of its five tandem
HMG-boxes in the process are presently unknown. Ligation-mediated circularization and supercoiling studies have suggested that HMG-box 1 alone may be sufficient (
25
). However, HMG- boxes 2 and 3 are known to be required to reconstitute the full DNA
binding affinity of wild-type (wt) xUBF (
23
) and to support
in vitro
transcription (
15
,
26
). Here we define the minimal protein and DNA requirements for the formation of
the Enhancesome.
To study the HMG-box requirements for the induction of a complete Enhancesome, three
truncation mutants of xUBF were produced, each containing an increasing number
of HMG-box domains (Fig.
1
B). The DNase I footprints of the mutant and wild-type xUBFs were found to be very similar (Fig.
1
C). [Those for xUBF, Nbox13 and Nbox1 have been previously described in some
detail (
23
) and are shown here to allow a comparison.] Each showed a clear footprint of ~20 bp immediately upstream and downstream of the
Xenopus
ribosomal RNA initiation site. However, some minor differences in the
footprints could be discerned. The two shorter mutants Nbox1 and Nbox12 gave
strong hypersensitivity at -1, +1 and protection to either side of this site as far as the
hypersensitive sites at -22 and +22, the previously defined boundaries for HMG-box 1 (
23
). However, these latter sites were especially weak in the case of Nbox12,
consistent with the idea that HMG-box 2 may protect bases upstream of -22 and downstream of +22. Wt xUBF and Nbox13 yielded a weaker
cleavage at -1, +1 than did Nbox1 and Nbox12, but gave stronger cleavages at the +22
and -22 sites. Thus, HMG-box 2 binding to the regions upstream of -22 and downstream of +22 may be modulated by the addition of
HMG-box 3. This may relate to the ESI findings described below. [The
hypersensitivity around -15, seen only when wt xUBF is bound, has been previously ascribed to the
acidic tail of xUBF, which is believed to fold back onto this region (
24
).]
ESI is ideally suited for visualizing DNA-protein complexes since, (i) the specimen does not have to be stained or
shadowed, (ii) it allows direct estimation of the mass of complexes and (iii) net phosphorous images localize the DNA component and allow the DNA content to be
estimated. Together, the mass information and the phosphorus content can reveal
stoichiometric relationships between protein and DNA. We have previously used
ESI to resolve the structure of the Enhancesome, showing it to contain a dimer of xUBF and a near complete loop of DNA (
24
) (see examples in Fig.
2
A-C). Therefore, we again used ESI to compare the DNA-protein structures formed by each of the xUBF mutants.
When the smallest mutant, Nbox1, was bound to the ribosomal enhancer DNA,
distinct complexes were seen (Fig.
2
D-F). The mass distribution of both the DNA and protein components of these complexes is shown in Figure
3
and summarized in Table
1
. They show that the great majority of complexes contained a dimer of Nbox1 (35 +- 9 kDa) and 62 +- 14 bp of DNA, only a little longer than the 40-44 bp we would expect from footprint data. The Nbox1
complexes were almost all associated with a kink or bend in the DNA. A similar
experiment with Nbox12 again gave complexes which were associated with a kink
or a bend in the DNA (Fig.
2
G-I). The Nbox12 complexes predominantly contained a dimer of the protein
moiety (27 of 36; 61 +- 8 kDa) and were associated with 102 +- 38 bp of DNA (Fig.
3
and Table
1
). Neither the Nbox12 nor the Nbox1 complexes showed any evidence of the type of
DNA looping associated with the Enhancesome. In the case of Nbox1, the
complexes also clearly did not contain sufficient DNA to form a loop of a
similar size to that observed in the wild type Enhancesome (Fig.
2
A-C and Table
1
).
Table 1
The numerical data obtained with each of the xUBF truncation mutants are given
The Nbox1 and Nbox12 complexes showed clear DNA kinking or bending. Estimates of
the degree of kinking could therefore be obtained by measuring the angle
between the incoming and outgoing branches of the DNA duplex. These data (Fig.
4
), showed a tight distribution of bend angles for both types of complex. In the
case of Nbox1, the data gave an average bend angle of 145 +- 24o, i.e., 73 +- 12o for each HMG-box 1. This bend angle is well within the range
of 30-130o measured for other HMG-box proteins (
18
,
20
). However, addition of HMG-box 2 in the Nbox12 mutant did not appear to greatly increase DNA bending,
an average bend of 160 +- 16o being measured for this mutant. This suggests that HMG-box 2 induces only a weak bending of the DNA of somewhere
between 0 and 27o (excluding the possibility of negative bend angles).
The DNA bends due to each HMG-box are shown in Figure
5
A as being in-phase, i.e., they all lie approximately in the same plane. However, for
this to occur the HMG-boxes must be precisely positioned such that they all bind to the same
face of the DNA. On B-form DNA, a spacing change of only one base pair between the binding sites
of adjacent HMG-boxes would already lead to a change of ~36o between the bends induced by these HMG-boxes. Larger spacing changes would lead to
proportionately greater angles. Such spacing changes would, therefore, be observed as very
significant pitch changes in the DNA path through the Enhancesome. What then
determines the positioning of the HMG-boxes on the target DNA? One possibility is that each HMG-box is directed to its site by DNA sequence preferences. However, it
has previously been shown that DNA-binding by xUBF is extremely sequence tolerant (
9
,
10
,
23
). A second possibility is that inter-HMG-box interactions enable the individual HMG-boxes of xUBF to position themselves correctly along the DNA.
In this case in-phase bending would be a property intrinsic to xUBF and would occur
equally on any DNA fragment, regardless of its sequence.
To decide between these alternatives, Nbox13 was bound to a 1 kb fragment of
bacterial DNA and analyzed by ESI in comparison with the enhancer DNA bound complex (Fig.
2
M-O). Mass analyses (Fig.
3
and Table
1
) (Nbox13-control DNA), showed that nearly 50% of the complexes (8 of 17) contained
a protein dimer (188 +- 35 kDa) and 122 +- 17 bp of DNA. Contour length measurements of the DNA confirmed
this latter figure (121 +- 21 bp). The complexes on the bacterial DNA showed a somewhat larger
dispersion of protein masses than those on the ribosomal enhancer DNA,
suggesting that their formation may be somewhat impaired. However, they
displayed no sequence (data not shown) and >50% showed a clear loop of DNA. The
diameter of this loop corresponded closely with that observed on the enhancer DNA (Fig.
2
J-L). Thus, it would appear that the formation of an Enhancesome is at
least in greater part independent of DNA sequence.
The formation of the Enhancesome complex, in which DNA is folded into a near 360o loop, is the very striking property of the polymerase I transcription factor xUBF (
24
). Here we have mapped the minimal protein domains required for this Enhancesome structure. The data
characterize the Enhancesome as containing a dimer of xUBF and 142 +- 30 bp of DNA wound into a single 350 +- 16o DNA turn of 16 +- 3 nm diameter. Formation of the structure
minimally requires a dimer of Nbox13, a C-terminal truncation mutant of xUBF containing only HMG-boxes 1-3 and the N-terminal dimerization domain (see Fig.
1
A). This segment of xUBF corresponds to that minimally required for activation
of polymerase I transcription (
15
,
26
). (Attempts to form complexes with a similar mutant in which this N-terminal dimerization domain was deleted were unsuccessful.) Reduction in
the number of HMG-boxes past 3 very significantly reduced the degree of DNA bending. A dimer
of the C-terminal truncation mutant Nbox1, containing only the N-terminal most HMG-box 1, induced a DNA kink of 145 +- 24o, while a dimer of Nbox12, containing HMG-boxes 1 and 2 bent the DNA by 160 +- 16o. Together the data suggest
that each of the first three HMG-boxes of xUBF induces an independent kink in the DNA and that these kinks
are phased in such a way as to be additive, i.e., they induce in-phase bending. Consistent with this, footprinting data have shown that the
two HMG-box 1s of an xUBF dimer bind to adjacent 20 bp segments of DNA and that
HMG-boxes 2 and 3 most probably occupy adjacent DNA sites [e.g. see Fig.
1
A and C and (
23
)]. Combining this information with the bend angles estimated for each HMG-box, the probable DNA path through the Enhancesome could be modeled (Fig.
5
A). This path is consistent with the dimensions of the Enhancesome.
Strong data exists suggesting that HMG-box domains bind the minor DNA groove and lie on the outside of the DNA
kink they induce (
19
,
20
,
35
). Hence, it is highly likely that the HMG-boxes of xUBF also bind around the outside of each DNA kink. The Enhancesome therefore probably resembles the structure modeled in Figure
5
A. Here, the six HMG-boxes of an Nbox13 dimer are shown bound around the outside of the
Enhancesome DNA loop. The model demonstrates the open nature of the Enhancesome
and it is evident that no protein-protein contacts could occur across the centre of the loop. This
emphasizes the extent to which the Enhancesome structure may rely on in-phase bending for its formation.
Bending by HMG-box 2 was found to be very small compared with that estimated for either
HMG-boxes 1 or 3 (box 1, 72 +- 12o; box 2, 0-27o; box 3, 95 +- 16o). Our data, however, leave open the
possibility that HMG-box 3 cooperates with box 2 in inducing full DNA looping. That is, within
the Enhancesome the bend angles per box may be more nearly equivalent.
Cooperativity between HMG-boxes could be exterted along the DNA duplex. However, the juxtaposition of the two HMG-box 3s within our model (Fig.
5
A) suggests that cooperativity might also occur via an interaction between these
boxes. That is via a clamping of the two ends of the DNA loop.
The exact positioning of each HMG-box along the DNA would appear to be a prerequisite for the formation of
the Enhancesome. However, xUBF and its individual HMG-boxes show little preference for any particular DNA sequence (
9
,
10
,
23
). When Nbox13 was bound on a randomly chosen bacterial DNA fragment,
characteristic Enhancesome complexes could be observed by ESI. Thus, at the
resolution of this technique, the HMG-boxes of the Nbox13 mutant are able to correctly position themselves on
DNA independent of its sequence. The model in Figure
5
A suggests that positioning could probably not result from direct contacts
between HMG-box domains and may, therefore, be due to a tethering effect of the short
intra-domain `linker' peptide sequences.
UBF is known to permit the binding of the TBP-complex, SL1, to the UCE and Core elements of the polymerase I promoter (
3
,
4
,
7
). What may then be the significance of the Enhancesome structure? The Nbox13
segment of xUBF, shown here to be sufficient for Enhancesome formation,
displays the full DNA binding affinity of xUBF (
23
). It also corresponds with the segment minimally required for
in vitro
transcription in
Xenopus
extracts (
26
). Data from human and
Xenopus
suggest that two distinct UBF dimers bind within the polymerase I promoter. One
of these dimers is associated with the UCE (
15
) and the other with the Core promoter element (
23
). A probable scenario for the arrangement of UBF binding along a stereotypical
vertebrate pol I promoter is shown in Figure
5
D. This arrangement of UBF would lead to the formation of two adjacent
Enhancesome structures within the promoter (Fig.
5
B and C). The resulting juxtaposition of the UCE and Core elements could then
provide a surface for the cooperative binding of a single or two interacting
SL1 complexes to these two promoter elements. This model also provides an
explanation of the requirement for stereo-specific alignment of the UCE and Core elements observed for the
mammalian, amphibian and even possibly the yeast polymerase I promoters (
36
-
38
). The close resemblance in DNA size, though not number of DNA turns, of the
Enhancesome to a chromatin Core nucleosome may also be more than a coincidence.
It is known that the
Xenopus
ribosomal enhancer chromatin loses its characteristic micrococcal nuclease
ladder on gene activation (
39
,
40
) (and unpublished data of Leblanc and Moss). However, it has also been shown
that the Core histones remain in contact with the enhancers via their N-terminal tails (
41
,
42
). The Enhancesome allows ample space to accommodate the Core histones within
the DNA loop. At the same time the diameter of the Enhancesome DNA loop would
of necessity prevent normal DNA-protein contacts with the globular Core of the histone octamer.
We wish to extend our thanks to M. Herfort for his excellent assistance. We also
thank Dr C. Read for the generous donation of the xUBF2 pGEX2T subclone of amino acids 110-189 (Box 1), and both Drs C. Read and C. Crane-Robinson for much technical advice and many useful discussions made
possible by a NATO collaborative research grant to C.R. and T.M. (no. 890637).
This work was supported by grants from the Medical Research Council of Canada
(MRC) to T.M. and D.P. B.-J. and the National Cancer Institute with funds from the Canadian Cancer
Society to D.P.B.-J. T.M. is an MRC of Canada Scientist and a member of the Centre de
Recherche en Cancérologie de l'Université Laval which is supported by the FRSQ of Québec. G.P. is supported by an award from the FCAR-FRSQ of Quebec.
REFERENCES
Return