ABSTRACT
The yeast TTAGGG binding factor 1 (Tbf1) was identified and cloned through its
ability to interact with vertebrate telomeric repeats
in vitro
. We show here that a sequence of 60 amino acids located in its C-terminus is critical for DNA binding. This sequence exhibits homologies
with Myb repeats and is conserved among five proteins from plants, two of which are known to bind telomeric-related sequences, and two proteins from human, including the telomeric
repeat binding factor (TRF) and the predicted C-terminal polypeptide, called orf2, from a yet unknown protein. We
demonstrate that the 111 C-terminal residues of TRF and the 64 orf2 residues are able to bind the
human telomeric repeats specifically. We propose to call the particular Myb-related motif found in these proteins the `telobox'. Antibodies directed against the Tbf1 telobox detect two proteins in
nuclear and mitotic chromosome extracts from human cell lines. Moreover, both
proteins bind specifically to telomeric repeats
in vitro
. TRF is likely to correspond to one of them. Based on their high affinity for
the telomeric repeat, we predict that TRF and orf2 play an important role at
human telomeres.
Telomeres are necessary to preserve the integrity of chromosomes during the cell
cycle by allowing their proper segregation during cell division, by preventing
their exonucleolytic degradation and end-to-end fusion, by positioning chromosomes within the nucleus and by
enabling complete replication of chromosomal ends (for reviews see
1
,
2
). In addition, yeast telomeres exert a position effect both on transcription (
3
,
4
) and on the timing of DNA replication (
5
). Finally, reports of nuclear movement mediated by telomeres during karyogamy
and meiotic prophase in fission yeast (
6
) suggest that telomeres may also be involved in meiotic chromosome movement.
The DNA sequence at telomeres is generally constituted of an array of tandem
repeats with clusters of G in the strand running 5' -> 3' towards the chromosome extremities, ending with a 3' overhang. The length of this repetitive DNA varies
among species and cell types. For example, the irregular (TG
1-3
)
n
telomeric sequence of
Saccharomyces cerevisiae
spans only few hundred base pairs, while the TTAGGG repeats at vertebrate
telomeres cover thousands of base pairs. A non-nucleosomal pattern of nuclease digestions is observed along the telomeric
DNA of ciliates (
7
,
8
) and yeast (
9
). In contrast, the terminal repeats from various vertebrate and invertebrate
species are arranged into regularly spaced nucleosomes smaller than bulk
nucleosomes (
10
). An absence of nucleosomes at the very end of human telomeres was suggested
from the diffuse nuclease digestion pattern observed with short human telomeres
(
11
). This particular telomeric chromatin is associated with various subnuclear
structures depending on species, cell type, stage in the cell cycle and
chromosomes (for a review see
2
). For example, mouse telomeres located on the long arm of chromosomes move from
the interior of the nucleus to the periphery between the G1 and G2 phases of
the cell cycle (
12
). During fission yeast meiosis telomeres are clustered near the spindle pole
body and the nucleolus (
13
). The dynamics of these associations are largely uncharacterized and may
require both stable and transient DNA-protein interactions.
In ciliate macronuclei the G-rich single-stranded tail is tightly bound to specific proteins that are believed to protect the
extremities from degradation (for a review see
14
). Such `capping proteins' may also exist in vertebrates, since a similar single-stranded binding activity was detected in
Xenopus
egg extracts (
15
) and several G-rich strand binding proteins have been reported in
Chlamydomonas
and
S.cerevisiae
(
16
,
17
). Most of the double-stranded part of
S.cerevisiae
telomeric DNA appears to be complexed with an array of multifunctional
repressor-activator protein 1 (Rap1), which distorts telomeric DNA as it binds
(
18
). Rap1 contacts non-DNA binding proteins to initiate propagation of transcriptionally repressed chromatin into adjacent non-telomeric sequences (
19
; for a review see
20
).
Nuclei from HeLa cells were isolated in polyamine buffers as previously
described in Gasser
et al.
(
27
). Soluble nuclear extracts (S100) from HeLa and Jurkat cells were prepared according to Dignam
et al.
(
28
). Heparin fractions from Jurkat cells were eluted by increasing KCl
concentration from a sulfopropyl 5PW column loaded with a 0.6 M KCl step
elution fraction from a heparin-agarose column as previously described (
29
). Metaphase chromosomes were isolated from HeLa cells blocked in mitosis (
30
). The purity of the chromosome preparation was checked by fluorescence
microscopy after DAPI staining (data not shown) and SDS-PAGE protein pattern as revealed by Coomassie staining (Fig.
1
A, lane 9). The total nuclear and chromosome extracts for gel electrophoresis
were obtained after digestion of the samples with microccocal nuclease for 1 h
on ice in 3.75 mM Tris-HCl, pH 7.4, 0.05 mM spermine, 0.125 mM spermidine, 20 mM KCl, 1 mM CaCl
2
and then adjusted to 2% SDS, 2% [beta]-mercaptoethanol and 10% glycerol. The amount of extract loaded per
well of SDS-polyacrylamide gel corresponds to ~10
6
nuclei.
We used the MalE expression system to produce the C-terminal part of Tbf1, hereafter called TBD, and its truncations. To
construct the plasmid pMalE-TBD, pMALtm-c2 DNA (purchased from Biolabs) was cut with
Eco
RI and ligated with the 800 nt
Eco
RI fragment from pCDS47 (
25
). pMalE-TBD encodes a 72 kDa hybrid protein, named E-TBD, consisting of maltose binding protein (MalE) lacking its
signal sequence, in order to be expressed into the cytoplasm, and the last 236
amino acids of Tbf1 separated by the recognition site of the protease factor
Xa. Plasmid p[Delta]1 (expressing a 67 kDa hybrid protein, named E-[Delta]1, in which amino acids 482-562 of TBD are missing) was constructed by deleting
the
Eco
NI-
Ear
I fragment from the coding region of TBD. Plasmid p[Delta]2 (expressing a 65 kDa hybrid protein, named E-[Delta]2, in which amino acids 404-468 of TBD are missing) was constructed by exchange
of the
Nco
I-
Stu
I fragment of p[Delta]1 for a PCR-amplified fragment corresponding to residues 326-404 of Tbf1. All constructs were checked by sequencing of the cloned fragments.
In order to obtain large amount of TBD for injection into rabbits, E-TBD was purified from bacterial cells in a procedure involving two
chromatographic steps.
Escherichia coli
strain DH5[alpha] transformed with plasmid pMalE-TBD was grown in 2* YT medium supplemented with 100 [mu]g/ml ampicillin to an OD
600
of 0.6, at which point IPTG was added to 2 mM. After a further 3 h cells were harvested and
lysed as previously described (
18
). The supernatant was applied to a heparin HyperD
TM
column (Sepracor SA). Most E-TBD, as monitored by Western blot analysis with anti-MalE antibodies (kindly provided by J.M.Clément), eluted in the 0.6 M NaCl fraction. This fraction was dialyzed against
50 mM Tris-HCl, pH 7.8, 0.2 M NaCl, 1 mM EDTA, loaded onto an amylose column and E-TBD protein was eluted with maltose as described (
31
), producing a highly purified protein, as checked by the presence of a single
band after heavily loading an SDS gel (data not shown). A similar purification
procedure was applied to proteins used for DNA binding assays, except for the
heparin column, which was omitted.
Band shift experiments were performed as described in Gilson
et al.
(
18
), except for the binding reaction buffer, which contained 20 mM KCl, 180 mM NaCl, 15 mM Tris-HCl, pH 7.4, 0.05 mM spermine, 0.125 mM spermidine, 1 mM dithiothreitol,
100 ng/[mu]l poly(dI[middot]dC), 100 ng/[mu]l bovine serum albumin. About 2 fmol of probe was used. All the
following probes were labelled at both ends with [[alpha]-
32
P]dATP using the Klenow enzyme. The HuTel
10
probe is a 110 nt
Eco
RI-
Hin
dIII fragment from pHuTel
10
containing 10 TTAGGG repeats cloned into the polylinker region of pGEM3Zf-. The HuTel
2.5
probe is a 110 nt
Eco
RI-
Hin
dIII fragment from pHuTel
2.5
containing the sequence TAGGGTTAGGGTTAGGG inserted in between the
Sac
I and
Kpn
I sites of pUC18.
The sequences of double-stranded oligonucleotides used as competitor are, for (TTAGGG)
2.5
, GTACCTAGGGTTAGGGTTAGGG annealed with TCGACCCTAACCCTAACCCTAG, and, for O.MYB,
GTACAACCTAACTGACACACAT annealed with TCGAATGTGTGTCAGTTAGGTT. Note that O.Myb
includes the sequence used for determination of the structure of c-Myb-DNA complexes (
33
).
Computer searches in sequence databases were performed using the BLAST algorithm
(
34
) and the e.mail servers available from NCBI. We used clustalW to produce the
multiple alignment (
35
). The phylogenetic tree was constructed using the neighbour-joining method (
36
). The percentage of difference between sequences was taken as an arbitrary
distance. Positions corresponding to gaps into the multiple alignment were not
taken into account for calculation of the percentage of difference. The telobox
consensus sequence was derived from multiple alignment of the eight members of
this family, retaining amino acids found at least four times in a given
position.
The sequences used for the multiple sequence alignment and tree construction
were: MybSt1 (
37
); IBP1 (
38
); BPF1(
39
); Rap1 (
40
); Tbf1 (
25
); human c-Myb (
41
); human A-Myb and B-Myb (
42
);
Xenopus
Myb1 and Myb2 (
43
); a set of open reading frames derived from partial cDNA sequences (EST) with
the GenBank accession nos Z26064 (orfA), D23805 (orfR1) and D22340 (orfR2);
orf1
and
orf2
, first identified from the assembly of R33191, R68526, R25990 and R70912 for
orf1
and T58911 and T11692 for
orf2
. Sets of primers containing built-in restriction sites were then designed in order to clone
orf1
and
orf2
.
Eco
RI sites were located at the 5'-end of the upstream primers (Pa and Pa'), whereas
Xba
I sites were located at the 5'-end of the downstream ones (Pb, Pb' and Pc') (Fig.
6
). After PCR amplification using the DNA of a HeLa cell cDNA library as template
(
44
) products of expected sizes were digested with
Eco
RI and
Xba
I and inserted in between the corresponding sites of pMALtm -c2 in order to fuse
malE
and
orf1
or
orf2
in-frame. Two independent clones emanating from each PCR experiment were
sequenced twice on both strands and found to have identical sequences (Fig.
4
). In each case, after IPTG induction and SDS-PAGE of a total protein extract, Coomassie staining revealed the
induction of an abundant protein whose migration corresponded to the expected
molecular weight for the hybrid protein (data not shown). These extracts were
used either for a Southwestern analysis or to transfer the hybrid protein onto
a nitrocellulose filter in order to purify TBD antibodies.
The procedures used are based on Miskimins
et al.
and von Kries
et al.
(
45
,
46
). After SDS-PAGE (8% polyacrylamide), without boiling the samples, gels were
equilibrated in blotting buffer (25 mM Tris base, 192 mM glycine, 10% methanol) and electrophoretically transferred onto nitrocellulose filters (BAS85, 0.45 mm; Schleicher & Schuell). The filters were incubated overnight at 4oC in 25 mM Tris base, 192 mM glycine. They were then incubated in binding
buffer (10 mM Tris-HCl, pH 7.4, 50 mM NaCl, 1 mM EDTA) supplemented with 5% non-fat milk for 90 min at room temperature with shaking, then for 90
min in binding buffer, 0.5% non-fat milk and again for 90 min at room temperature in binding buffer, 0.5%
non-fat milk, end-labeled probe (~300 fmol/ml, corresponding to 10
6
c.p.m./ml), 5 [mu]g/ml poly(dI[middot]dC), 50 [mu]g/ml
E.coli
DNA and competitor plasmid DNA as indicated (Figs
1
and
4
). Filters were washed for 20 min in 2 l ice-cold binding buffer supplemented with 0.5% non-fat milk. Buffer was changed several times. Then filters were
quickly dried on 3MM paper and exposed for autoradiography on Kodak X-Omat films and amplifying screens at -70oC.
The HuTel
10
probe is described above. The YTel probe is a 130 nt
Eco
RI-
Hin
dIII fragment from pYtCA-1X (
47
) that contains 80 nt of TG
1-3
repeats. pHuTel
700
was constructed by inserting between the
Eco
RI and
Sma
I sites of pUC18 plasmid DNA a mixture of PCR products digested with
Eco
RI and
Stu
I generated in a template-free reaction (
48
) using primers GCGGAATTC(TTAGGG)
8
and GAAGGCCTC(TAACCC)
8
. Sequencing of the resulting plasmid DNA revealed the presence of only TTAGGG
repeats over the 700 nt of the inserted DNA in pHuTel
700
.
Since the yeast Tbf1 protein specifically recognizes the vertebrate telomeric repeat (TTAGGG)
n
, we hypothesized that a human telomeric protein might contain a DNA binding
domain related to that of Tbf1. To explore this, a polyclonal rabbit antiserum
was raised against the C-terminal 236 residues of Tbf1 (amino acids 326-562; see Materials and Methods), which were known to include the
protein DNA binding domain (called TBD;
25
). Western blot analysis of a total yeast protein extract revealed with the TBD
antiserum (Fig.
1
A, lane 1) shows a single band corresponding to the size of Tbf1 (65 kDa). When
a protein extract from the same yeast strain expressing a hybrid protein
between the transactivation domain of Gal4 and TBD is probed with this serum an
additional band corresponding to the expected size for the Gal4-TBD hybrid protein is seen (data not shown). These results demonstrate
that the serum is monospecific for Tbf1 in yeast.
The antibodies directed against TBD were affinity purified using TBD as the
antigen and were used in blotting experiments with human nuclear proteins. In
total protein from isolated HeLa nuclei the anti-TBD antibodies recognized a major band with an electrophoretic migration
identical to that of Tbf1 and an upper band of lesser intensity (Fig.
1
A, lane 2, Coomassie staining, and lane 8). The major band was also detected
among total proteins from isolated metaphase chromosomes (Fig.
1
B, lane 1). Using the secondary antibodies alone neither of the two bands were
detected (data not shown). We ruled out that the human nuclei or chromosomes
were contaminated with yeast by probing total human nuclear extracts with an
antiserum directed against Rap1, an abundant yeast nuclear protein. No cross-reacting protein was detected (data not shown). A soluble extract prepared from isolated nuclei
(S100) also exhibits a major signal at the level of Tbf1 closely migrating with
an upper band of lesser intensity (Fig.
1
B, lane 4); note that in this experiment SDS-PAGE was run for longer in order to improve the resolution of the two bands. We conclude
that two human nuclear polypeptides with apparent molecular weights close to
that of Tbf1 are specifically recognized by antibodies directed against the
Tbf1 DNA binding domain.
To determine whether the human proteins detected by TBD affinity-purified antibodies also bind human telomeric DNA a Southwestern assay was
performed on the same protein samples using a duplex DNA probe containing 10 TTAGGG repeats (named HuTel
10
). Remarkably, the major (TTAGGG)
10
binding activity in total nuclear extracts migrates with an apparent molecular
weight of 65 kDa, exactly co-migrating with the Tbf1-related proteins (Fig.
1
A, compare lane 2 with lanes 3 and 4). Since this activity is completely eliminated by the presence of a 20-fold excess of the specific competitor pHuTel
700
DNA (Fig.
1
A, lane 5), but not by a 200-fold excess of non-specific pUC18 DNA (Fig.
1
A, lane 3), the binding appears to be highly specific for telomeric repeats. A similar competition pattern was obtained using a DNA probe
including 40 TTAGGG repeats (data not shown). Since (TTAGGG)
n
sequences, like most telomeric repeats, are TG-rich on one strand (
1
), we tested whether or not the 65 kDa band would bind the irregular (TG
1-3
)
n
sequence of
S.cerevisiae
telomeric DNA. With the double-stranded yeast telomeric probe YTel we were unable to detect any
interaction with either 65 kDa protein (Fig.
1
A, lanes 6 and 7), lending further support to the high selectivity of this human
protein for (TTAGGG)
n
sequences. A similar band co-migrating with the major immunoblotting activity was detected in a
Southwestern experiment with total mitotic chromosome proteins (Fig.
1
B, lane 2).
Performing a similar Southwestern assay on human proteins from a soluble nuclear
extract, two proteins exactly co-migrating with the doublet protein detected by immunoblotting bind the
HuTel
10
probe (Fig.
1
B, lanes 3 and 4). We reproducibly failed to detect the upper band in lysed
total nuclei, both by immunoblotting and by Southwestern assays (Fig.
1
A, lanes 2-4 and longer exposures of these autoradiographs; data not shown). This
may be due to abundant proteins in whole nuclei, that are less abundant in
soluble extracts, which interfere with either transfer to the filter or with
DNA binding of the upper band. Treatment of the soluble extract with calf intestinal phosphatase did not modify the migration of either
band of the doublet, as revealed by either Southwestern or immunoblotting
experiments (data not shown). Thus the upper band is unlikely to correspond to
a phosphorylated form of the lower band.
We next asked whether the protein bands at 65 kDa detected by immunoblotting
really correspond to the proteins detected by Southwestern assay. To do this we
subfractionated the soluble nuclear extract from Jurkat cells through several
chromatographic steps and tested all samples by both immunoblotting and Southwestern assay. In
the initial extract two closely migrating proteins ~65 kDa are recognized by anti-TBD antibodies (Fig.
2
, lane 1), like those detected in a similar nuclear extract of HeLa nuclei (Fig.
1
B, lane 4). After a heparin-agarose column the two proteins, as revealed by both DNA binding and
immunoblotting, co-elute at 0.6 M KCl (data not shown). The 0.6 M heparin fraction was then
loaded onto a sulfopropyl 5PW column, where again immunoblotting and
Southwestern assay detect two closely migrating proteins that elute between 200 and 240 mM KCl in fractions F15-F17 (Fig.
2
, lanes 2-9); in the Southwestern assay the two closely migrating bands in these
fractions are more easily seen in a lower exposure of the autoradiograph (data
not shown). The binding specificity for (TTAGGG)
n
DNA of the activities recovered in fraction F16 was confirmed by competition
experiments (data not shown). F16 is still a mixture of several polypeptides,
as revealed by SDS-PAGE followed by Coomassie staining, and no major polypeptide at 65 kDa
is visible (data not shown).
Since the two proteins revealed by Southwestern assay after two successive
chromatographic steps are indistinguishable by SDS-PAGE analysis from the two revealed by immunoblotting, it is very likely
that both assays identify the same proteins. Thus we conclude that there exist
two human polypeptides (the p65 doublet; Fig.
1
B) that bind specifically to human telomeric DNA and that are recognized by
affinity-purified antibodies directed against the Tbf1 DNA binding domain.
In order to map the epitopes of Tbf1 that are shared with the p65 doublet, we
tested antibodies purified against a series of truncated forms of TBD. A TBD
lacking 64 central amino acids ([Delta]1) and a TBD lacking 80 terminal amino acids ([Delta]2) were expressed and purified from
E.coli
cells as a hybrid protein with the bacterial maltose binding protein MalE (E-[Delta]1 and E-[Delta]2, left part of Fig.
3
A). Antibodies affinity purified against E-[Delta]1 detected the p65 doublet present in fraction F16 poorly, while
they bind perfectly a full-sized E-TBD protein expressed in bacterial cells (right part of Fig.
3
A). In contrast, when purified against E-[Delta]2 they efficiently recognized both p65 and E-TBD (right part of Fig.
3
A). These results demonstrate that most of the epitopes shared between TBD and
the p65 doublet are localized within amino acids 404-468 of Tbf1. Interestingly, this region contains an essential part of the
Tbf1 DNA binding domain, since its deletion abolished DNA binding. This is demonstrated both by band shift assay
with highly purified hybrid proteins (Fig.
3
B, compare lanes 6-8, wild-type E-TBD, with lanes 12-14, E-[Delta]1) and Southwestern assay with total
bacterial extracts containing either E-TBD or E-[Delta]1 (Fig.
4
A). In contrast, a TBD with the 80 amino acid C-terminal deletion still binds telomeric DNA in a sequence-specific manner (Fig.
2
B, lanes 9-11; data not shown). We conclude that the Tbf1 DNA binding domain is
found between positions 326 and 482 and includes, between positions 404 and
468, most of the epitopes shared with the p65 doublet.
We attempted to identify p65 by searching human protein sequences homologous to
the sequence (amino acids 404-468) of Tbf1 in various databases using the BLAST algorithm (
34
). We found that the sequence corresponding to amino acids 406-457 exhibits homology to the DNA binding domain of Myb proteins (Fig.
5
; data not shown). These DNA binding domains are often constituted by three
imperfect tandem repeats (R1, R2 and R3), with R2 and R3 making specific
contacts with the cognate DNA sequence (
33
). An alignment between Tbf1 and the third repeat of a typical member of the Myb protein family, namely human c-Myb, is shown in Figure
5
A. Although the homology is partial, many of the highly conserved residues
important for R3 are present within the Tbf1 sequence. In particular, two out
of the three expected tryptophan residues are present. However, in contrast to
most known Myb proteins, the Tbf1 sequence has only one such repeat.
Intriguingly, Tbf1 does not specifically bind DNA sequences recognized by
various known Myb proteins. This is demonstrated in Figure
3
B, where increasing amounts of a double-stranded oligonucleotide carrying either a DNA sequence recognized by most
known Myb proteins (
49
,
33
) (lanes 3-4) or 2.5 repeats of TTAGGG (lanes 1-2) were added as competitor to the band shift assay. Similarly, no
competition was observed using two other DNA sequences recognized by plant
MYB.Ph3 (
50
; data not shown).
Since Tbf1 does not bind a typical Myb DNA site and contains only a single Myb
repeat (see above), Tbf1 may belong to a particular subfamily of Myb proteins.
Indeed, among the Myb proteins revealed in the BLAST search for proteins
homologous to the Tbf1-like Myb repeat two have a single Myb repeat and bind specific DNA
sequences related to telomeric repeats, namely the maize
Shrunken
initiator binding protein IBP1 (
38
) and the parsley BoxP binding factor BPF1 (
39
). The IBP1-DNA complex at the
Shrunken
promoter covers an exact plant telomeric repeat (AGGGTTT) (
38
) and BPF1 binds a series of sequences rich in G
n
T
m
motifs (
39
). Both proteins share an almost identical single Myb repeat (one mismatch out
of 54 amino acids) located at their C-termini (Fig.
5
B) which is implicated in DNA binding (
38
,
39
). This suggests that, in addition to their function in transcriptional
regulation, these proteins may play a role in plant telomere physiology.
Comparison of the Myb repeat with the DNA binding domain of the well-characterized yeast telomeric factor Rap1 shows that Rap1 also contains
two regions of homology with the Myb box (
40
).
To further test the possibility that all members of the `Tbf1 family' bind
telomeric DNA (see above) we investigated the DNA binding properties of the two
human open reading frames derived from partial cDNA sequences, namely orf1 and
orf2. The DNA coding for orf1 and orf2 was PCR-amplified from a HeLa cDNA library as described in Materials and Methods.
After cloning in-frame with the
malE
gene synthesis of the E-orf1 and E-orf2 hybrid proteins was checked by Coomassie blue staining of
total bacterial extracts analysed by SDS-PAGE. An abundant induced protein of the expected molecular weight was
confirmed by immunoblotting using anti-MalE antibodies (data not shown). The sequence recovered by PCR largely confirmed the expected sequence from
an EST assembly (Fig.
6
A and B), with the presence of a stop codon in-frame with the
Myb
-containing open reading frames. Since this stop codon was present within
the sequence of the PCR primers Pb and Pb', we also amplified
orf2
with a primer located downstream of Pb' (Pc', Fig.
6
B); sequence determination through Pb' confirmed the presence of the stop codon for
orf2
. The amino acid sequences of orf1 and orf2 outside the Myb-related region are not related (Fig.
6
C), confirming that orf1 and orf2 reflect portions of two different proteins
having a Myb/Tbf1-related domain at their extreme C-terminus. Intriguingly, for all the characterized members of the `Tbf1 family', i.e. Tbf1,
IBP1, BPF1, orf1 and orf2, the single Myb repeat is located at the extreme C-terminus of the protein.
Crude extracts from
E.coli
cells expressing either E-orf1 or E-orf2 were subjected to a Southwestern analysis with HuTel
10
as probe. In both cases a telomeric DNA binding activity co-migrates with the hybrid protein (Fig.
4
A). As a control extracts from bacteria expressing only MalE or the hybrid
protein E-[Delta]1, which does not bind (TTAGGG)
n
in band shift assays (Fig.
3
B), failed to exhibit any telomeric DNA binding activity and an extract containing E-TBD exhibits a binding activity at the level of the hybrid protein (Fig.
4
A). This shows that, like TBD, orf1 and orf2 bind telomeric DNA sequences. The
specific binding of E-orf1 and E-orf2 to (TTAGGG)
n
sequences was further analysed by specific and non-specific competition experiments. Up to a 200-fold molar excess of non-specific competitor DNA (pUC18) over the probe did not affect
binding of either E-orf1 or E-orf2 or E-TBD (Fig.
4
B). In contrast, a similar molar increase of the specific competitor pHuTel
700
DNA, containing 700 nt of TTAGGG repeats inserted into pUC18, greatly reduced binding of the three hybrid
proteins (Fig.
4
C, lanes 4-12). It is worth noting that pHuTel
60
DNA, containing 60 nt of TTAGGG repeats, was ~40 times less efficient in competition as compared with pHuTel
700
DNA, further indicating that the competition is dependent upon the number of
TTAGGG repeats (Fig.
4
C, compare lanes 1-3 and 7-9). Finally, the binding of either orf1, orf2 or TBD is unaffected
by the presence of up to a 1400-fold molar excess of O.Myb oligonucleotide over the probe, showing that,
like TBD, orf1 and orf2 do not exhibit specific binding to a typical Myb DNA
site. Overall, these results demonstrate that orf1 and orf2, like TBD,
specifically bind human telomeric DNA and that the minimal telomeric DNA
binding domain of the proteins from which orf1 and orf2 are derived is found
within the 111 C-terminal amino acids of orf1 and the 63 C-terminal amino acids of orf2.
The fact that orf1 and orf2 bind telomeric DNA sequences and share homologies
with amino acids 404-466 of Tbf1 strongly suggests that these open reading frames may be
identical to the p65 doublet. When TBD antibodies were affinity purified
against either E-orf1 or E-orf2 they recognized roughly equally p65 and E-TBD (right part of Fig.
3
A). Since TBD antibodies purified against MalE alone do not react with p65 and E-TBD (data not shown), we can conclude that orf1 and orf2 are both
immunologically related to Tbf1.
This study has identified a particular Myb-related protein motif that appears to be specialized for specific
interaction with duplex telomeric DNA. This motif is present in proteins from
yeast (Tbf1), plants (IBP1 and BPF1) and in partial sequences of human open reading
frames (orf1 and orf2). We propose to name this motif the `telobox'. Other
putative members of the telobox family include open reading frames from rice
(orfR1 and R2, Fig. 5B) and from
Arabidopsis
(orfA, Fig.
5
B). The sequence of orf1 is identical to the C-terminal part of the human telomeric protein TRF, whose sequence was
published after this work was completed (
23
). Since orf1/TRF is located at chromosome ends
in vivo
(
23
; unpublished results), this protein is expected to play an important role at
telomeres. This may also be true for orf2, which binds TTAGGG
in vitro
with the same affinity as orf1/TRF (Fig.
4
). Overall, these results strongly support the existence of two human telomere-associated proteins sharing a telobox at their C-terminus. The respective role of each protein in telomere physiology
remains to be determined.
A telobox consensus was derived from the multiple alignment presented in Figure
5
B, revealing a bipartite structure with a central region that is variable in
length and sequence (Fig.
7
A). Roughly 30% of residues are identical between the N-terminal 27 residues of the telobox consensus and R2/R3 of c-Myb, while only 10% are identical with the two c-Myb repeats within the C-terminal 19 residues (Fig.
5
A). In particular, the C-terminal VDLKDKWRT sequence of the telobox consensus shows limited
homology with the sequence of R2 and R3, while it is highly conserved among the
telobox members, including orf1 and orf2 (Fig.
5
B). Interestingly, the corresponding regions of R2 and R3 contain the residues
that establish specific contacts with the bases of the Myb consensus site, as
revealed by NMR analysis (
33
) (Fig.
7
A). This suggests that this telobox motif might also be crucial for specific
telomeric sequence recognition. In this respect it is worth noting that the
sequence of a vertebrate telomeric repeat is distantly related to the Myb
binding site consensus (Fig.
7
B). Thus it is tempting to speculate that the various Myb motifs, through slight
modifications in their DNA contacting residues, can bind different but related
sequences. This is in agreement with the fact that Tbf1, orf1/TRF and orf2
recognize telomeric DNA repeats, while they are unable to specifically bind
typical Myb DNA sites (Figs
3
B and 4D).
The telobox constitutes the major part of the telomeric DNA binding domain, at
least for orf1/TRF and orf2, which contain little more than the telobox motif.
However, in IBP1 a stretch of basic residues following the telobox motif was
shown to also be necessary for efficient DNA recognition (
38
). Thus the telobox flanking residues might also be required to stabilize or to
properly fold the telobox or to contact additional DNA sites.
The level of homology between distantly related telobox sequences is high enough
to allow interspecific immunological cross-reactivities. For example, antibodies directed against the yeast Tbf1
telobox specifically interact with two human telobox peptides (orf1 and orf2;
Fig.
3
A). Furthermore, we have shown that antibodies directed against the telobox of
Tbf1 almost exclusively detect two human nuclear proteins of ~65 kDa (the p65 doublet), both of which specifically bind telomeric DNA
sequences in a Southwestern assay (Figs
1
and
2
). At least one of the p65 polypeptides is likely to correspond to TRF, which
has a similar apparent molecular weight (
23
). Whether orf2 corresponds to the other p65 polypeptide or to another protein
remains to be determined.
The fact that Rap1 does not fall into the telobox family based on our
phylogenetic tree analysis (Fig.
5
C) suggests that other types of Myb-related domains may also be used for binding telomeric repeats. In
addition to the sequence divergence between Rap1 and telobox proteins, the Rap1
DNA binding domain contains two Myb-related motifs instead of one (
40
). It is worth noting that the sequence of the
S.cerevisiae
telomeric DNA (TG
1-3
)
n
is quite different from the sequences found in a wide phylogenetic range of
eukaryotes, including the vertebrate TTAGGG sequence (
53
). This suggests that during evolution a new telomeric repeat sequence might
have been added to the existing one, requiring recruitment of an ancient Rap1
precursor for efficient telomere maintenance. The presence of TTAGGG-like repeats at the junction between yeast telomeric repeats and the
interior of chromosomes may thus represent `relic' sequences from an ancient
telomere (
26
). The conservation of Tbf1 in yeast is probably due to non-telomeric functions, which remain uncharacterized but may be linked to
regulation of transcription. Like Rap1, which acts both as a structural
component of yeast telomeres and as a transcriptional regulator (
54
), both IBP1 and BPF1 were first identified as promoter binding elements. It
will be revealing to examine other members of the `telobox' family to elucidate
whether or not the presence of a Myb-related DNA binding domain and involvement in transcriptional regulation
represent universal characteristics of telomere binding proteins.
We thank D. Shore and R.K. Moyzis for gifts of plasmids, R. Brent for providing
the HeLa cDNA library, W. Saurin for his help in the phylogenetic tree
construction, J. M. Clément for his help during the amylose-heparin chromatography, G. Lombart-Platet for generously sharing the sulfopropyl fractions and
C. Brun for critical reading of the manuscript. C.E.K. thanks the Association
Française de Lutte contre la Mucovisidose (AFLM) for her long-term fellowship. S.M.G. thanks the Swiss National Science
Foundation and the Human Frontier Program for continued support. E.G. thanks
EMBO for his long-term fellowship in the Gasser Laboratory, as well as the Ligue contre le
Cancer, ARC, GREG, Region Rhône-Alpes and AFLM since moving to Lyon. The nucleotide sequence data reported in this paper will appear in
the EMBL, GenBank and DDBJ Nucleotide Sequence Databases under the accession
nos X93511 for orf1 and X93512 for orf2.
+
Present address: Department of Pathology, McMaster University, Hamilton, Ontario
L8N 3Z5, Canada
REFERENCES
Return