ABSTRACT
Mammalian ribosomal genes are flanked at their 5
'
and 3
'
ends by terminator sequences which are recognized by the transcription
termination factor TTF-I. The occurrence of the same binding site upstream and downstream of the
gene raises the possibility that TTF-I can interact with both sequences simultaneously and thus brings the
terminator in the vicinity of the gene promoter by looping out the pre-rRNA coding sequence. To test this model, we have examined the ability of
TTF-I to oligomerize and found that both full-length and N-terminally truncated versions of TTF-I form stable oligomeric structures. At least two
domains of TTF-I located within the 184 N-terminal and 445 C-terminal amino acids, respectively, mediate the self-association of several TTF-I molecules. In support of the looping model, TTF-I is capable of linking two separate DNA
fragments via binding to the target sites. This result indicates that in
addition to its function in transcription termination, TTF-I may serve a role in the structural organization of the ribosomal genes
which may be important for maintaining the high loading density of RNA
polymerase I on active rRNA genes.
In eukaryotes, transcription of the genes that code for ribosomal RNA (rDNA)
accounts for up to 80% of cellular RNA which is being synthesized at any
instant in a rapidly growing cell. This high transcriptional activity is
brought about by maximal density of RNA polymerase I (Pol I) together with an
amplification of the number of transcription units. In mammals, ~200 copies of rDNA per haploid genome are required to synthesize the more
than a million ribosomes per generation that are needed to maintain the
translational capacity of the new daughter cells. In most species, rDNA is
arranged in tandem head-to-tail repeats in which a transcribed region alternates with an
intergenic spacer region. Consistent with the role of rRNAs in ribosome
structure and function, the gene regions that code for 18S, 5.8S and 28S rRNAs
are highly conserved. The intergenic spacer, on the other hand, exhibits a
pronounced heterogeneity, both in length and sequence. However, even though the sequence of regulatory elements that govern Pol I
initiation and termination vary considerably between species (for reviews see
1
,
2
), the overall structural organization of the rDNA repeats is similar. Several
types of regulatory elements are located in the intergenic spacer, including
(i) the gene promoter at the 5' end of the pre-rRNA coding region, (ii) a transcription terminator immediately
upstream of the gene promoter, (iii) enhancer elements that stimulate
transcription, (iv) one or more spacer promoters, and (v) terminator elements
at the 3' end of the pre-rRNA coding region. Specific transcription factors bind directly or
via protein-protein interactions to these regions and thus promote the
synthesis of faithfully initiated and terminated pre-rRNA.
Eukaryotic ribosomal transcription units are flanked both at the 5' and 3' side by one or more terminator elements. In mouse, the 18 bp
terminator motif, termed `Sal box' because it contains a restriction site for
Sal
I, is repeated several times (T
1
-T
8
) downstream of the 3' end of the pre-rRNA coding region and has been shown to mediate transcription
termination
in vivo
and
in vitro
(
3
-
5
). Alterations in the `Sal box' that reduce binding of the interacting factor
TTF-I (for
T
ranscription
T
ermination
F
actor) also impair transcription termination. There is marked sequence
divergence between terminator elements from different organisms, such as
mammals, frog and yeast. The molecular mechanism of Pol I transcription
termination, however, is probably similar or even identical in these diverse
species. All characterized Pol I terminators function in only one orientation
and bind a termination factor which presumably contacts the elongating RNA
polymerase (for review, see
6
). The cDNAs for murine and human TTF-I have been cloned and deletion analysis has revealed functionally
distinct domains of the protein (
7
,
8
). Interestingly, the DNA binding activity of recombinant TTF-I (TTF-I
p130
) has been found to be masked in the intact protein (
9
). Removal of the N-terminal part of TTF-I, on the other hand, greatly augments DNA binding. These findings
suggested that the N-terminus of TTF-I may inhibit DNA binding via intermolecular protein-protein
interactions. Consistent with this idea, we found that the N-terminal 184 amino acids of TTF-I can form stable oligomers in solution and repress DNA binding when
fused to a heterologous DNA binding domain (
10
).
The fact that binding sites for the termination protein are present both
upstream and downstream of the rDNA transcription unit suggests a functional
linkage between transcription termination and initiation. A model has been
proposed in which each rDNA transcription unit forms a loop which juxtaposes
the promoter and the terminator element (
11
,
12
). Thus, Pol I molecules having terminated at the downstream terminator could be
transferred directly from the 3' end of the gene to the promoter of the adjacent rDNA unit without
entering the free pool. The finding that a sequence motif that is almost
identical to the downstream terminator elements is also located adjacent to the
rDNA promoter suggests that simultaneous binding of a sequence-specific protein to both the upstream and downstream terminators may
connect the 5' and 3' end of the rDNA and therefore mediate DNA looping. In the loop
structures that are supposed to be formed, interaction between the upstream and
downstream terminators of the same or adjacent transcription units can be
juxtaposed. Planta and collaborators have suggested that REB1p, with indentical
binding sites near the promoter and the 3' end of the rRNA operon, is causally involved in loop formation (
12
). The experimental data presented in this study strongly support this model. We
demonstrate that TTF-I, the murine homologue of yeast REB1p can self-associate and form oligomeric structures both in solution and when
bound to DNA. The intermolecular interactions between different TTF-I molecules, in turn, enable the factor to interact simultaneously with
two separate DNA fragments bearing a TTF-I binding site. The results are compatible with the hypothesis that TTF-I may link the proximal and distal part of the rDNA transcription
units as distinct loop structures.
Expression vectors containing histidine-tagged TTF-I
p130
, TTF[Delta]N185, TTF[Delta]N323 and TTF[Delta]N445 in pRSET (Invitrogen) were described by Evers
et al
. (
7
). For expression in baculovirus- infected Sf9 cells,
Nde
I-
Hin
dIII fragments containing the histidine-tag and TTF-I sequences from the pRSET constructs were cloned into pBacPAK9
(Clontech). GST-TTF[Delta]N323 was generated by cloning a
Bam
HI (blunt)/
Eco
RI fragment from pRSET-TTF[Delta]N323 into pGEX-3X (Pharmacia).
Proteins were expressed by infecting 2.5 * 10
8
Sf9 cells with recombinant baculovirus. The cells were harvested after 48 h,
rinsed in PBS, resuspended in 3 vol of lysis buffer (50 mM HEPES-KOH, pH 7.8;
300 mM KCl; 5 mM MgCl
2
; 1 mM PMSF; 1 [mu]g/ml leupeptine), and were lysed by sonification followed by addition of
0.5% NP-40 and centrifugation. The supernatant was incubated with NTA-agarose
beads (Quiagen) for 30 min at 4oC in the presence of 1 mM imidazole. The beads were washed with 20 column
volumes of buffer 1 (50 mM HEPES-KOH, pH 7.8; 300 mM KCl; 5 mM MgCl
2
; 0.5% NP-40; 1 mM imidazole; 1 mM PMSF; 1 [mu]g/ml leupeptine), 20 vol of buffer 2 (same as buffer 1 with 1 M KCl)
and 20 vol of buffer 3 (same as buffer 1 with 10 mM imidazole). Proteins were
eluted with 20 mM HEPES-KOH, pH 7.8; 100 mM KCl; 5 mM MgCl
2
; 200 mM imidazole; 1 mM PMSF; 1 [mu]g/ml leupeptine and dialysed against buffer AM-100 (20 mM Tris-HCl, pH 7.9; 5 mM MgCl
2
; 100 mM KCl; 0.1 mM EDTA; 20% glycerol; 2 mM DTE).
To determine the size of native TTF-I by gel filtration, 25-50 [mu]l aliquots of
35
S-labeled TTF-I derivatives synthesized by
in vitro
translation (Promega) were centrifuged and passed over a Superdex200 FPLC
(HR10/30, Pharmacia) column at a flow rate of 0.5 ml/min in buffer AM-100 without glycerol (20 mM Tris-HCl, pH 7.9; 5 mM MgCl
2
; 100 mM KCl; 0.1 mM EDTA; 2 mM DTE). The fractions were precipitated with
trichloroacetic acid, analyzed by SDS-PAGE and autoradiography, and the amount
of TTF-I in individual fractions was quantified with a PhosphorImager.
TTF[Delta]N323 fused to glutathione S-transferase (GST-TTF[Delta]N323) was expressed in
Escherichia coli
BL21(DE3) and purified on glutathione-Sepharose beads as specified by the
manufacturer (Pharmacia). 50 [mu]l assays contained 2 [mu]g of fusion protein bound to 10 [mu]l of glutathione-agarose beads and 2-5 [mu]l of
35
S-labeled full-length or mutant forms of TTF-I in buffer AM-100. The reactions were incubated for 30 min at room
temperature and washed three times with buffer AM-100. The washed beads were boiled in sample buffer and the released
proteins were electrophoresed on 10% SDS-polyacrylamide gels.
20 [mu]l reactions containing TTF-I and 5 fmoles of a
32
P-labeled 246 bp PCR fragment covering rDNA sequences from -232 to +14 (relative to the transcription start site) were
incubated for 15 min on ice in binding buffer (12 mM Tris-HCl, pH 8.0; 85 mM
KCl; 5 mM MgCl
2
; 0.1 mM EDTA; 1 mM DTE; 8% glycerol; 2 ng/[mu]l BSA; 4 ng/[mu]l phage [lambda] DNA cut with
Hae
III; 0.1% NP-40), and protein-DNA complexes were separated by electrophoresis on 4%
polyacrylamide gels in 0.5* TBE buffer (50 mM Tris-borate, pH 8.3; 1.3 mM EDTA) at 4oC and 10 mA. For competition, a double-stranded `Sal box' oligonucleotide (SB; upper strand 5'-GATCCTTCGG
To assay TTF-I binding to two different DNA fragments, a 160 bp PCR fragment containing
the terminator element T
2
(
9
) was generated using a biotinylated primer. The fragment was attached to
streptavidin-coated magnetic beads (Dynal) according to the manufacturers
specifications. Ten fmoles of immobilized DNA, 10 fmoles (40 000 c.p.m.) of
32
P-labeled `Sal box' oligonucleotide (SB or SB*) and TTF-I were incubated for 30 min at 30oC in binding buffer containing 0.5% NP-40. Protein-DNA complexes were isolated in a magnetic field,
washed in 50 [mu]l buffer AM-100, eluted with 15 [mu]l of loading buffer (10 mM Tris-HCl, pH 8.0; 5 mM EDTA; 1% SDS;
30% glycerol; 0.01% bromphenolblue; 0.01% xylene cyanol) and analyzed on a 12%
native polyacrylamide gel.
In a recent study we have compared DNA binding and termination activity of
recombinant full-length TTF-I (TTF-I
p130
) with two deletion mutants lacking 184 and 322 N-terminal amino acids (TTF[Delta]N185 and TTF[Delta]N323). These studies revealed that the DNA binding of TTF-I
p130
is at least one order of magnitude lower than that of the deletion mutants
indicating that the N-terminus represses the interaction of TTF-I with DNA (
10
). Moreover, on glycerol gradients both full-length TTF as well as a polypeptide encompassing the N-terminal 320 amino acids (TTF1-320) sedimented much faster than expected for a monomeric
protein. These and other data suggested that the N-terminal domain of TTF-I has the potential to oligomerize with itself and that
oligomerization of TTF-I may influence its DNA binding activity.
To investigate whether the ability of TTF-I to form stable oligomers in solution was mediated exclusively by the N-terminus or whether the C-terminal part of TTF-I is also able to interact with itself, we determined
the native size of TTF[Delta]N445, a mutant harboring the C-terminal DNA binding domain. For this, radiolabeled protein,
synthesized by
in vitro
translation, was subjected to gel filtration on a Superdex200 column. Figure
1
shows the distribution of TTF[Delta]N445 in individual column fractions as analyzed by SDS-PAGE and
autoradiography. Similar to previous studies showing that TTF-I
p130
forms oligomers in solution (
10
), a significant part of the N-terminally truncated mutant TTF[Delta]N445 also eluted in the void volume or in fractions which represent
molecular sizes larger than expected for a monomeric 43 kDa protein.
Thus, not only the very N-terminus, but also the C-terminal part of TTF-I can mediate self-association of several TTF-I molecules. The same result was obtained with
highly purified TTF-I derivatives that were expressed in baculovirus-infected Sf9 cells and purified by affinity chromatography (data not
shown) indicating that the intermolecular interaction of several TTF-I molecules is due to oligomerization of TTF-I and is not dependent on other factors present in the translation
lysates.
Next we investigated whether TTF-I would also form multimeric complexes when bound to DNA. In the
experiment shown in Figure
3
, increasing amounts of TTF[Delta]N185 purified from baculovirus-infected insect cells were incubated with a labeled rDNA fragment
containing one TTF-I binding site and the resulting protein-DNA complexes were analyzed by electrophoresis. In the presence of
20 fmoles of TTF-I, the DNA probe was quantitatively converted into a distinct DNA-TTF-I complex (complex C1, lane 2) which exhibits a lower
electrophoretic mobility than free DNA. Significantly, when the amount of TTF-I was increased to 500 fmoles, complex C1 was converted into a more slowly
migrating complex (C2, lane 3).
This finding, together with the observation that TTF-I forms oligomers in solution, suggests that complex C2 most likely
contains several TTF-I molecules. If this assumption was correct, then addition of a `Sal box'
oligonucleotide which competes for TTF-I binding should convert complex C2 into complex C1. The competitions
shown in Figure
3
(lanes 4-7) support this view. At saturating amounts of TTF-I (20 fmoles), an excess of `Sal box' oligonucleotide containing the TTF-I target sequence (SB), but not a mutant oligonucleotide
(SB*), efficiently competed for TTF-I binding to the labeled DNA probe (lanes 4 and 5). If the competitions
were performed at high concentrations (500 fmoles) of TTF-I, addition of wild-type `Sal box' oligonucleotide, but not the mutant, prevented
formation of complex C2. Consistent with a stepwise dissociation of the complex
C2, complex C1 and distinct intermediates were observed (lanes 6 and 7). No
free probe was generated, because of the high amounts of TTF-I in the assay required to produce complex C2. Significantly, the same
complexes migrating between complex C1 and C2 were observed if intermediate TTF-I concentrations, i.e. between 20 and 500 fmoles, were used (data not
shown). This result suggests that at high molar ratios of TTF-I to DNA several TTF-I molecules bind simultaneously to the `Sal box' target sequence.
Although these data do not allow definite conclusions about the stoichiometry
of TTF-I binding, the observation that complex C2 is more prominent than the two
intermediate complexes indicates that complex C2 is a multimeric complex,
presumably a tetramer, which is more stable than dimers or trimers.
The ability of TTF-I to form multimers when bound to DNA raises the possibility that TTF-I could connect two DNA segments containing TTF-I binding sites. This possibility is particular intriguing
because it would be compatible with the `ribomotor' model which proposes that
each rDNA transcription unit forms a loop which places the gene promoter and
terminator into close proximity (
11
). Our data suggest that TTF-I may be the
trans
-acting factor that bridges the 5' and the 3' end of the transcription unit by binding simultaneously to
the upstream and the downstream terminator(s). If this model is correct, then
TTF-I should be able to link spatially separated DNA fragments. To test this
idea, a `bridging assay' was designed, as illustrated in Figure
4
A. In this assay, a DNA fragment containing a TTF-I binding site was immobilized on magnetic beads and incubated with a
radiolabeled `Sal box' oligonucleotide in the absence and presence of TTF-I
p130
. Protein-DNA complexes formed on the immobilized DNA were isolated by magnetic
separation and analyzed for the presence of labeled oligonucleotide. As shown
in Figure
4
B, association of labeled `Sal box' oligonucleotide with bead-bound DNA was dependent on TTF-I. Increasing the input of TTF-I
p130
increased the amount of labeled oligonucleotide (SB) in the bead-bound fraction (lanes 1-6). This interaction was dependent on TTF-I binding to its target sequence, because a labeled mutant
oligonucleotide (SB*) that is not recognized by TTF-I did not associate with the immobilized DNA (lanes 7-9). This result indicates that DNA-bound TTF-I can link two separate DNA segments. In control
reactions that lack TTF-I (lanes 10 and 11) or contain a non-immobilized DNA fragment (lanes 12-14) no significant levels of radiolabeled `Sal box'
oligonucleotide were found to be associated with the beads. Together, this
finding demonstrates that TTF-I may tether different DNA molecules. However, the overall amount of
`sandwich'
complexes containing two `Sal box'-containing DNA fragments, hold together by TTF-I, is low. Whether this is due to experimental manipulation or due
to the instability of this kind of complexes is not known.
To delineate the region of TTF-I which tethers separate DNA molecules, various TTF-I mutants were tested for their ability to physically link the
immobilized rDNA fragment and the radiolabeled `Sal box' oligonucleotide. Purified baculovirus-expressed full-length TTF-I (TTF-I
p130
) as well as the deletion mutants TTF[Delta]N185, TTF[Delta]N323 and TTF[Delta]N445 were analyzed in the `bridging assay' described above
(Fig.
5
). Significantly, TTF[Delta]N185 and TTF[Delta]N323 were as active as intact TTF-I in tethering the two DNA fragments, indicating that
deletion of 322 N-terminal amino acids did not affect the simultaneous interaction of TTF-I with spatially separated DNA fragments. In contrast, mutant TTF[Delta]N445 was inactive in this assay (lanes 1, 3 and 5). This is
an interesting observation, because sequences between amino acids 323 and 445
have been shown to play an important role in both transcription termination (
7
) and in TTF-I directed chromatin remodeling (
13
). Thus, the failure of TTF[Delta]N445 to link physically separated DNA fragments supports the view that
the C-terminus on its own, including the DNA binding function of TTF-I and the ability to oligomerize, is not sufficient for TTF-I function, but requires sequences between amino acids 323 and
445.
In order to complete the transcription cycle, RNA polymerase must undergo
termination which includes cessation of elongation and the release of both the
terminated RNA chains and RNA polymerase from the template. The common view of
reinitiation is that RNA polymerase needs to be released in order to be
recruited by preinitiation complexes to start a new transcription cycle.
However, the finding that in yeast the rDNA terminator maps within a DNA region
that enhances transcription initiation suggested a functional linkage between
termination and initiation. A model has been proposed which implies that each
rDNA transcription unit forms a loop which may channel polymerases directly to
the promoter after termination, thus bypassing the pool of free Pol I molecules
(
11
). This is an attractive model because it reveals a possible mechanism by which
the high level of rDNA transcription is accomplished. A looping mechanism would
maintain a high loading density of the rDNA transcription unit by ensuring
efficient recycling of Pol I from the 3' tail to the 5' head of active genes. In support of this model, micrographs of
chromatin spreads from
Bombyx mori
and
Drosophila
tissue culture cells show active rDNA transcription units as loops separated by
intergenic spacers (
14
).
In this study we provide experimental support for this model. Our data suggest
that TTF-I, besides its function in termination of Pol I transcription, serves a
role in the structural organization of active rDNA transcription units. The
following lines of evidence argue that interaction between several TTF-I molecules bound to either the promoter proximal or distal terminator
elements may connect the 5' and 3' end of the gene. First, the position of the upstream
transcription terminator T
0
with respect to essential promoter elements has been conserved. In mouse, rat,
human,
Xenopus laevis
and
X.borealis
, T
0
is located ~200 bp upstream of the transcription initiation site and therefore, may be
part of the promoter itself. T
0
has been shown to stimulate transcription initiation
in vivo
to some extent (
15
-
17
). This positive effect on transcription has been interpreted to be the result
of shielding the promoter from polymerases that read through from spacer
promoters, thereby inactivating or `occluding' productive initiation complexes
(
16
,
18
). However, consistent with the looping model, the upstream terminator also
stimulates transcription initiation by a mechanism which is dependent on the
helical alignment between the terminator and the rDNA promoter (
19
).
A second argument for TTF-I connecting distant rDNA regions is the observation that TTF-I forms oligomeric structures. We demonstrate that the ability of
TTF-I to associate with itself is not restricted to the N-terminal part, but that an additional oligomerization domain is also
contained within the C-terminal region including amino acids 445-833. Interestingly, the two oligomerization domains appear to be
functionally different. The N-terminal domain (which resides between amino acids 1 and 184) has
previously been shown to form stable oligomers in solution and to repress the
DNA binding activity of full-length TTF-I (
10
). TTF[Delta]N445, the mutant that specifically binds DNA but is inactive in
transcription termination (
7
), also forms oligomeric complexes in solution. However, this mutant fails to
link two physically separated DNA molecules. Apparently, oligomerization of TTF-I in solution
per se
is not sufficient for linking separate DNA segments. This result underscores
the importance of the central part of TTF-I including amino acids from 323 to 445 in functions other than DNA
binding. We propose that this central part of TTF-I, together with the C-terminal DNA binding domain, is not only essential for transcription
termination (
7
) and chromatin remodeling (
13
), but also plays a crucial structural role in organizing the rDNA transcription
units and spacer regions.
Stable protein-protein mediated DNA loops may provide a general mechanism by which
distant DNA sites modulate gene expression. Multimeric structures, such as homo-multimers of E2 dimers and of Sp1 tetramers, frequently assemble at loop
junctures (
20
-
23
). One example of how homomeric protein oligomerization may affect gene
expression via a DNA looping mechanism is the tumor suppressor protein p53.
Natural p53 binding sites placed adjacent to TATA elements effectively
stimulate transcription by p53 (
24
). p53 exists as tetramers and multiples of tetramers in solution (
25
-
27
). In a model promoter containing multiple copies of the consensus sequence, p53
has been found to assemble oligomeric complexes by a novel mechanism which
stacks tetramer on top of tetramer. Morover, stacked oligomers link separated
binding sites via DNA loops and promote transcriptional enhancement
in vivo
(
28
). By analogy to TTF-I, p53 can assemble oligomers by two distinct domains, a C-terminal tetramerization domain and a non-tetrameric oligomerization domain that loops separated
consensus sites by protein-protein interactions.
The possibility that polymerase `hand-over' from the end of one gene to another rDNA promoter may be responsible
for the high polymerase density seen on ribosomal genes has previously been
tested in microinjected oocytes (
29
) as well as in cultured kidney cells (
30
). The studies revealed that high rates of transcription initiation do not
depend upon polymerase passing from one repeat to the next. We have also
performed a series of
in vitro
transcription experiments to find out whether the presence of both an upstream
and a downstream terminator would increase the initiation frequency on
artificial ribosomal minigenes. We consistently observed transcriptional
enhancement by TTF-I (data not shown). However, this enhancement was not due to communication
between the promoter-proximal terminator and the downstream TTF-I binding site, but was brought about by the downstream terminator
alone. These and other experiments indicate that TTF-I stimulates transcription, presumably by facilitating the reinitiation
reaction
.
Therefore, a final proof whether or not RNA polymerase I is `handed over' from
the downstream terminator to the gene promoter is still lacking. Nevertheless,
our data suggest that TTF-I may be causally involved in maintaining a loop structure of the rDNA
transcription units. Whether the interaction between the upstream and
downstream terminators is mediated exclusively by TTF-I, or whether it involves additional proteins that may anchor the rDNA to
the nucle(ol)ar matrix in a highly ordered, linear fashion remains to be
investigated.
We thank Stephen Mason for advice and helpful discussions. This work was
supported in part by the Deutsche Forschungsgemeinschaft (SFB 228) and the Fond
der Chemischen Industrie.
Sander, E.E, Mason, S.W., Munz, C. and Grummt, I. (1996)
Nucleic Acids Res.
,
24
, 3677-3684.
*To whom correspondence should be addressed. Tel: +49 6221 423 423; Fax: +49
6221 423 404; Email: I.Grummt@DKFZ-Heidelberg.de
REFERENCES
Return