ABSTRACT
CTCF belongs to the Zn finger transcription factors family and binds to the
promoter region of c-
myc
. CTCF is highly conserved between species, ubiquitous and localised in nuclei.
The endogenous CTCF migrates as a 130 kDa (CTCF-130) protein on SDS-PAGE, however, the open reading frame (ORF) of the CTCF cDNA
encodes only a 82 kDa protein (CTCF-82). In the present study we investigate this phenomenon and show with
mass-spectra analysis that this occurs due to aberrant mobility of the CTCF
protein. Another paradox is that our original cDNA, composed of the ORF and 3
'
-untranslated region (3
'
-UTR), produces a protein with the apparent molecular weight of 70 kDa
(CTCF-70). This paradox has been found to be an effect of the UTRs and sequences
within the coding region of the CTCF gene resulting in C-terminal truncation of CTCF-130. The potential attenuator has been identified and point-mutated. This restored the electrophoretic mobility of the
CTCF protein to 130 kDa. CTCF-70, the aberrantly migrating CTCF N-terminus
per se
, is also detected in some cell types and therefore may have some biological
implications. In particular, CTCF-70 interferes with CTCF-130 normal function, enhancing transactivation induced by CTCF-130 in COS6 cells. The mechanism of CTCF-70 action and other possible functions of CTCF-70 are discussed.
CTCF is a novel 11 Zn-finger transcription factor, present in nuclear extracts as a protein with
an apparent MW of 130 kDa (CTCF-130) on SDS-PAGE. It was discovered for its ability to bind to an unusually
long 45 bp GC-rich sequence containing three regularly spaced repeats of the core
sequence CCCTC within the chicken
c-myc
promoter (
1
). However, its ability to recognise diverged sequences by employing different
combinations of zinc fingers has recently been demonstrated (
2
). Since deregulation of the
c-myc
proto-oncogene is causally related to oncogenesis (
3
,
4
), identification and characterisation of factors regulating its transcription
may provide insight into molecular mechanisms of normal and aberrant
c-myc
expression (
3
-
6
). Thus, CTCF features were further investigated (
1
,
7
) and its cDNA cloned (
8
).
CTCF has been found to be expressed in different chicken tissues in multiple
forms (130, 97, 80, 73, 70 and 55 kDa) as detected by anti-CTCF antibodies (
8
). In various cell types the CTCF protein is encoded in a 4.1 kb mRNA, with the
longest open reading frame (ORF) (2184 bp) of the cDNA predicting a 728 amino
acid protein of 82 kDa. This contradicts the observed MW of 130 kDa. The
discrepancy between the theoretical (82 kDa) and apparent (130 kDa) MWs of the
CTCF protein products could be due to a number of possibilities: (i) there is
an additional exon in the primary transcript, which is missing in our cDNA
possibly generated by alternative splicing (
9
-
12
); (ii) post-translational modifications (
13
-
19
) could change the mobility of the CTCF protein; (iii) particular amino acid
composition (
20
-
23
) could lead to CTCF anomalous electrophoretic migration. These possibilities
are discussed in this paper.
Of the other multiple forms of CTCF expressed, in this study we investigated the
appearance of the 70 kDa protein. Previously, it was found that when the
original CTCF cDNA containing the entire ORF and the 3'-untranslated region (3'-UTR) was expressed in cells, it produced the 70 kDa
protein product, but not 82 or 130 kDa (
8
). We investigated the following options for this discrepancy: (i) post-translational processing (
24
-26) or protein splicing (
27
); (ii) transcription and (iii) translation attenuation (
28
-
30
).
In this study we show that that the discrepancy between the 82 and the 130 kDa
MWs of the CTCF protein was due to anomalous migration in the SDS-PAGE (mechanism 3) and that the N- and C-terminal domains participate in this anomaly. We also show
that the CTCF-70 protein represents a truncated version of the full length CTCF protein
(CTCF-130), corresponding to the N-terminus, and is probably generated by premature termination of
translation of the CTCF mRNA at a specific site within the CTCF coding region.
This depends on the sequences within the UTRs and the coding region. Silent
nucleotide substitution (i.e. mutation without changing the CTCF amino acid
sequence) at the potential attenuator site inside the ORF restored the size of
CTCF-70 to 130 kDa.
Additionally, we demonstrate that the 70 kDa protein, corresponding to the
abnormally migrating CTCF N-terminal domain, enhanced transactivation induced by the full length wild
type (wt) CTCF in COS6 cells.
Plasmids pEEc and pBBc were constructed by inserting the full length 3.786 kb
Eco
RI-
Eco
RI fragment from plasmid p900-911 and the 3.647 kb
Eco
RI-
Eco
RI fragment from p900 (
8
), respectively, into the eukaryotic expression vector pSG5, controlled by early
SV40 promoter (Stratagene). pHHc was obtained by deleting a part of the 3'-UTR of pBBc with
Hin
dIII, resulting in a 2.855 kb insert (Fig.
1
A). `
Pst
I truncated' versions of all three plasmids (pBBc[Delta]PP, pHHc[Delta]PP and pEEc[Delta]PP) were obtained by `in-frame'
Pst
I-
Pst
I deletion of 840 bp within the coding sequences. pBBc, pHHc, pBBc[Delta]PP, pHHc[Delta]PP and pEEc[Delta]PP
myc
-tagged derivatives were obtained by introducing a 294 bp (
Hin
cII-
Bam
HI) fragment, encoding six
myc
tag epitopes, from the plasmid p6
Myc
Tags-D (
31
) into the
Msc
I site at the 3'-terminal end of the CTCF ORF (Fig.
1
A). The pBBc-mut plasmid, containing point-mutations within the region of a proposed attenuator, was generated
by PCR using two oligonucleotides beginning at nucleotide 385: 5'-TGATGATGGAGCACCTGGAT-3' and nucleotide 836: 5'-GAAGCTTACTCTT
COS6 cells were grown at low (40-50%) density in Dulbecco's modified Eagle's medium supplemented with 10%
foetal bovine serum, and transfected with 1-10 [mu]g of pBBc, pHHc or pEEc plasmids by the calcium phosphate
precipitation method with 2* DNA precipitation buffer following the manufacturer's instructions (`5
Prime-3 Prime', Inc
R
, CP Laboratories, UK). Cell extracts were made 48 h post-transfection. For co-transfection assays 1.25 * 10
5
COS6 cells were plated in 16 mm dishes and 1 [mu]g of reporter (p90TK Luc) and 0.5-2.5 [mu]g of expressor (pHHc or pBBc) were introduced with Lipofectamine (2.5 [mu]l/dish) following the manufacturer's instructions (Gibco BRL). The total
amount of the DNA transfected was adjusted to 3.5 [mu]g with the `empty' pSG5 vector. Cells were harvested 16-18 h after transfection. Luciferase assays were performed with a
`Dual-Light
TM
' kit (Tropix, USA) following the manufacturer's instructions using a TD-20e luminometer (Turner, USA).
Total cell lysates were treated with DNAse I in the presence of 5 mM MgCl
2
, electrophoresed on 10% SDS-PAGE gels, transferred onto Immobilon P membranes (Millipore, Bedford,
MA) by semi-dry blotting, and probed with either the chicken-specific anti-CTCF antibody Ab2 (
8
) or monoclonal 9E10 anti-
myc
tag antibody (
34
). CTCF or
myc
-tag-fusion proteins were visualized by the standard ECL procedure with
an ECL detection system (Amersham).
Aliquots of 1 [mu]g of the pBBc, pEEc and pHHc DNA were translated
in vitro
using the TNT
R
coupled reticulocyte lysate system, according to the manufacturer's protocol
(Promega) from the T7 promoter present in the pSG5 expression vector. The
synthesised proteins were resolved on 10% SDS-PAGE gels, and blotted onto Immobilon-P membrane where necessary.
The glutathione-S-transferase (GST)-fusions were expressed in TG-1 cells and purified on glutathione-Sepharose as previously described (
33
), then to homogeneity by anion exchange FPLC using a mono-Q column (Pharmacia) and eluted at ~0.2 M NaCl with a salt gradient (0.5-0.1 M) in a 20 mM Tris-HCl buffer pH 8.0. The purified GST-fusions were cleaved with thrombin, then treated
sequentially with benzamidine beads (Sigma) and glutathione- Sepharose (Pharmacia) to remove the protease and the GST, respectively.
The molecular weight of the cleaved proteins was determined by comparing their
mobility with that of Rainbow molecular weight markers (Amersham) after
electrophoresis on 10% polyacrylamide-SDS gels.
The cleaved proteins were also sequenced from their N-termini in order to confirm their identities. The true molecular weights
were determined by matrix-assisted laser desorption mass spectrometry (MALD-MS) using the Finnigan Lasermat 200 mass spectrometer. The N- or C-portions of the CTCF protein were isolated by FPLC on a
MonoQ column (Pharmacia), the protein samples were reduced with 20 mM
dithiothreitol, desalted by HPLC on a C4 reverse phase column (Brownlee) with
acetonitrile 0.1% trifluoroacetic acid. After chromatography the sample was
dried under vacuum, taken up in 10% methanol, and then 0.5 ml was mixed with
0.5 ml of sinopinic acid matrix (10 mg/ml) solution in 70% acetonitrile on a
slide following the manufacturer's instructions. The mass spectrometer was
calibrated with bovine serum albumin (Sigma,
M
r
= 66 430) for the uncleaved proteins and carbonic anhydrase II (Sigma,
M
r
= 29.023) for the protein of molecular weight <30 000 Da. Each spectrum was determined at least three times and the mean
values calculated.
To explain the discrepancy between the predicted 82 and 130 kDa MWs of the CTCF
protein, we first considered mechanism 1, i.e., the possibility of an
additional exon in the primary transcript which is missing in the cDNA. All our
investigations, including screening various cDNA libraries, RACE and RT-PCR on various total and polyA
+
mRNA preparations, failed to identify any additional coding sequences which
could produce a 130 kDa form of CTCF protein. Therefore mechanism 1 was ruled
out and the remaining options examined: mechanism 2, i.e. post-translational modifications of the protein and mechanism 3, i.e. CTCF-82 anomalous electrophoretic migration. In order to investigate
these options, three parts of the CTCF cDNA (encoding the N-terminus, Zn-finger domain and C-terminus) were subcloned into pGex2TK vector (N- and C-termini) or pGex3T (Zn-finger domain). Recombinant GST-fusion proteins were expressed in
Escherichia coli
, purified and analysed by mass-spectrometry. Mass-spectrometry analysis did not reveal any post-translational modifications (mechanism 2). As shown in Table
1
, the experimental molecular weights of the GST-N fusion and N- and C-termini alone, determined by MALDI-MS, correspond well with their theoretical MWs. However,
SDS-PAGE analysis demonstrated anomalous electrophoretic mobility of the GST-fusion proteins and the N- and C-termini. GST-Zn migrates as a 98 kDa protein in contrast to
its predicted MW of 81.4 kDa (20% anomalous migration, Fig.
2
A, lane 4). This fusion protein is rather unstable, giving rise to the products
of degradation appearing as a smear after the staining with the anti-GST-antibodies. GST-C migrates as a 70 kDa protein, in contrast with its predicted
size of 44 kDa (59% aberrant migration, Fig.
2
A, lane 2). GST-N appears as a predominant band of ~97 kDa in contrast to its predicted size of 53.8 kDa (79% of aberrant
migration, Fig.
2
A, lane 3), with two minor bands most likely resulting from degradation. It is
worth noting that the presence of the GST-Zn fusion protein of ~98, GST-N of 97 and GST-C of 70 kDa has been consistently detected in independent
experiments involving staining with Coomassie and anti-GST-antibodies, whilst the positions of the minor bands varied.
Following the GST removal the C- and N-termini migrate aberrantly at 35 and 70 in contrast to their
predicted 18 and 28 kDa (94 and 160% anomalous migration, respectively, Fig.
2
B, lanes 2 and 3). Thus the differences between the theoretical molecular
weights and the apparent molecular weights of the C- and N-terminal domains are due to anomalous electrophoretic migration of
the polypeptides (mechanism 3). This would in turn account for the anomalously
high apparent molecular weight of the intact CTCF protein.
A
B
Analysis of sequences within the 3'- and 5'-UTRs affecting the size of the CTCF protein products,
revealed that addition of only 106 bp (
Eco
RI-
Sma
I fragment, Fig.
1
A) of the 5'-UTR to the 3.672 kb cDNA (pBBc, Fig.
1
A), restored the size of the protein expressed
in vivo
and
in vitro
to 130 kDa (pEEc, Fig.
1
A). This fragment is very GC-rich (73% of GC-content, ref.
8
) and its effect is specific. Thus, replacing it with other fragments (either GC- or AT-rich) did not change the size of the protein (70 kDa) produced from
these hybrid cDNAs (data not shown). On the other hand, addition of a 792 bp of
the 3'-UTR (
Hin
dIII-
Eco
RI fragment, pEEc, Fig.
1
A) to the plasmid pHHc-130, resulting in pBBc, reduced the apparent size of the protein made from
130 to 70 kDa. However, removal of this fragment from pEEc (pHHE, Fig.
1
A) did not change the size of the protein produced (130 kDa, data not shown).
Thus, only particular combinations of the UTRs with sequences inside the coding
region are responsible for the production of CTCF-70.
It has been noticed that the size of the 70 kDa protein corresponds to the size
(68 kDa) of the CTCF N-terminal domain migrating abnormally (Fig.
1
A, black arrow). This, together with the observation that the C-terminal domain is missing in CTCF-70 (Fig.
4
), suggests the existence of a potential attenuation site within the coding
region close to the end of the N-terminal domain with a likely position inside the sequence containing
poly(A) stretches (Fig.
1
B and ref.
8
). To locate the region in pBBc where premature termination of translation
occurs the in-frame
Pst
I-
Pst
I deletion, which removes the putative attenuatior site, was obtained in all
three plasmids, pBBc, pHHc and pEEc (Fig.
1
A). As shown in Figure
5
A (lanes 1-3), after transfection in COS6 cells, all three plasmids carrying the
deletion produced a protein of the same size (97 kDa). This finding suggested
the presence of the potential attenuator in pBBc within this deleted region. To
further prove that this deletion does prevent translation termination and the C-terminal portions of all the proteins produced are intact, we employed the
same `
myc
-tag' approach as described above. For this purpose a sequence encoding six
myc
-tag epitopes was cloned in frame into the
Msc
I site (the end of the Zn-finger domain) of all three plasmids (see also Materials and Methods and
Fig.
1
). As shown in Figure
5
A (lanes 4-6) the three plasmids after transfection in COS6 cells and Western-immunoblot with the chicken-specific CTCF antibodies, produce proteins of the same size (~110 kDa which corresponds to the size of the
Pst
I-
Pst
I deleted plasmids: 97 kDa plus the
myc
-tag component, 11 kDa). When the membrane was stripped of the anti-CTCF antibodies and reprobed with 9E10, the same bands appeared
positive (Fig.
5
B, lanes 4-6). The multiple bands of smaller size observed in all our
myc
-tagged proteins (including pHHc
myc
, Fig.
4
A and B, lane 2) are not products of premature termination, but most likely
resulted due to proteolysis of the
myc
-tagged proteins containing foreign sequences and therefore unstable
because: (i) they are detected only with the 9E10 anti-
myc
-tag antibodies and not with the anti-N-terminal CTCF antibodies; (ii) they are specific only to the
myc
-tagged plasmids; and (iii) the same fragments can be seen in all three
myc
-tagged plasmids (Fig.
5
B, lanes 4-6). Thus the potential terminator of translation is very likely to be
located within the
Pst
I-
Pst
I fragment and its removal prevents premature termination.
To investigate whether the poly(A) sequences within the
Pst
I-
Pst
I region are involved in attenuation, silent point-mutations were introduced into pBBc, creating pBBc-mut, such that they did not change the amino acid sequence, but at
the same time disrupted the poly(A) stretches (Fig.
1
B). This alteration in the sequence restored the size of the protein expressed
in vivo
to 130 kDa (Fig.
5
C, lane 3) suggesting the involvement of these poly(A)-stretches in premature termination.
The ability of one single ORF to produce
two alternative products both,
in vivo
and
in vitro
, raises a question of the CTCF-70's functional significance. CTCF-70 can be detected as a minor form in some cell types, but in bursa
cells CTCF-70 represents ~25% of the total CTCF proteins (
8
). As CTCF-70 is correctly located in the nucleus (data not shown), efficiently
expressed and corresponds to the CTCF N-terminal domain, one could expect that it may affect the normal function
of 130 kDa protein. In fact, biological role of terminal fragments has been
previously reported for some transcription factors. In particular, the C-terminal domain of p53 (
35
) or B-
myc
which represents the N-terminal domain of the
myc
protein (
36
) can affect transactivation caused by the full length protein. To investigate
CTCF-70 ability to interfere with CTCF-130 we employed a standard transactivation assay in COS6 cells as a
model system. In these cells co-transfection with pHHc and p90TKLuc [a reporter plasmid containing the
dimeric CTCF binding site from the chicken
c-myc
(
1
,
8
), inserted in front of a TK Luciferase reporter gene] resulted in
transactivation of the luciferase gene (Fig.
6
A). Transactivation was dependent on the input of the expressor pHHc. As
expected, no transactivation was observed with pBBc in the same assay (Fig.
6
B). However, CTCF-70 enhances the ability of CTCF-130 to transactivate the reporter when both expressors are present
in COS6 cells (Fig.
6
C).
The major purpose of the experiments described above was to clarify the
relationship between the native CTCF protein and those encoded and produced by
the cloned cDNA. Evidence has been provided that: (i) the 130 kDa product is
encoded in the 2184 bp ORF which predicts a 82 kDa protein; (ii) this
discrepancy is due to an anomalous migration of the protein in the SDS-PAGE gel; (iii) the 70 kDa CTCF protein is a truncated version of CTCF-130 representing its N-terminal domain; and (iv) the 3'- , 5'-UTRs and sequences within the CTCF
coding region affect the size of the expressed protein.
The results in this paper demonstrate that the 82 kDa CTCF protein, encoded in
the longest ORF of 2.184 kb, migrates as a protein with an apparent molecular
weight of 130 kDa in SDS-PAGE, whether the recombinant protein was produced
in vitro
or
in vivo
. This corresponds to the apparent size of the endogenous native CTCF protein.
We have also shown that the recombinant proteins are recognised by anti-CTCF antibodies demonstrating that the endogenous and recombinant proteins
are identical. The aberrant mobility of the protein can be mainly accounted for
by the electrophoretic mobility properties of the N- and C-terminal domains. The anomalous electrophoretic migration of the CTCF C-terminus may be explained by its high negative charge (26%
acidic amino acids). Anomalous electrophoretic migration of negatively charged
proteins is quite common and has been observed with proteins such as
papillomavirus 16 E7 protein (
20
) and
E.coli
ams
protein (
37
). We also attempted to characterise the CTCF N-terminal domain aberrant migration. However, when pEH (Fig.
1
A) was deleted up to the
Bam
HI site and the resulting plasmid pBE expressed in COS6 cells, a 180% anomaly
(21 kDa apparent versus 7.5 kDa theoretical MWs) was still observed. The reason
for the anomalous migration of the N-terminal region remains unclear since there is no preponderance of particular types of amino acids.
We have previously reported (
8
) that the original CTCF-expressing vectors based on the pBBc construct produced a nuclear protein with apparent molecular weight of 70 kDa
in vivo
and
in vitro
(Fig.
3
). The results of this paper demonstrate that: (i) CTCF-70 is a legitimate, not transient, protein product from pBBc since
proteins with higher molecular weight from this construct have never been
observed, and (ii) its size reduction is due to the loss of the C-terminal part of the CTCF protein. There are at least three possible
mechanisms responsible for the appearance of the shorter protein: (i) post-translational processing/splicing of the full length protein; (ii)
premature termination of transcription or RNA degradation; and (iii) premature
termination of translation of the CTCF mRNA. Mechanism 1 can be ruled out since
we were unable to detect a 130 kDa precursor from the pBBc plasmid, after
conducting a variety of studies such as pulse-chase experiments,
in vitro
translation, and investigation of a stable cell line producing the 70 kDa
protein (
8
, and unpublished data). Furthermore, proteolysis is considerably unlikely, as no
additional bands were detected in extracts from pBBc
myc
-transfectants probed with 9E10 anti-
myc
antibodies (Fig.
4
B), and protein splicing is not applicable either, as the entire C-terminal (and, presumably, the Zn finger) domains are missing. Mechanism 2
can be excluded on the basis that with Northern blot analysis, we were unable
to detect any additional RNA bands smaller than the 3.7 kb band, which
corresponds to the full-length CTCF RNA in various tissues and cell lines including a stable cell
line producing the 70 kDa protein (
8
). Mechanism 3 is quite plausable, as three lines of evidence indicate that CTCF-70 is produced by premature termination of translation: (i) the entire C-terminal part of the CTCF-70 protein is missing as shown by the
myc
-tag approach (Fig.
4
); (ii) deletion of the region incorporating the putative attenuator prevents
premature termination of translation (Fig.
5
A and B); and (iii) point-mutation of the potential attenuator site restores the size of CTCF-70 to 130 kDa (Fig.
5
C).
Our investigations demonstrate that only particular combinations of the UTRs
with sequences inside the coding region are responsible for the production of
CTCF-70. Analysis of the 3'-UTR 792 bp sequence and the coding region revealed the
presence of poly(T) and poly(A) stretches, respectively (
8
). This implies the possibility of an attenuator produced by complementary
interaction between the poly(T) and poly(A) stretches. The molecular basis of
the production of the CTCF-70 is not known. However, based on data presented in this study and
significant information accumulated in the understanding of the general
translational termination processes (
38
-
42
) we have proposed a model for this mechanism. This model is the basis of
ongoing and future investigations of this study and, therefore, seems relevant
to this discussion. The involvement of the 5'-UTR and 3'-UTR in the regulation of translation has been
convincingly established (
30
,
43
,
44
). Much less is known about elements within coding regions which could
contribute to this process. However, it has been shown that sequences within
the coding region can affect the mRNA translation efficiency (
45
,
46
) and translationally coupled mRNA degradation (
47
). Our model of the molecular mechanism of premature termination involves both
the UTRs and the coding region. In the case of pBBc composed of the ORF and 3'-UTR, the formation of the stem-loop secondary structure between the poly(T) sequences
within the 3'-UTR and poly(A) stretches within the putative attenuator, could
block the ribosome's passage through the CTCF mRNA. As a result a truncated
protein product of 70 kDa is produced. In the case of pEEc, the additional GC-rich 5'-UTR sequences may cause a ribosome `stalling' at the 5'-end of the CTCFmRNA. This `stalling', in turn,
can lead to destabilisation of the `downstream' CTCFmRNA stem-loop structure allowing the ribosome to progress. pHHc, producing CTCF-130, is missing parts of its 5'- and 3'-UTRs. Thus, the pHHc mRNA may not have a
developed secondary structure so that translation can be performed without
spatial difficulties. This model is supported by the finding that the specific
point mutations within the region of the potential attenuator in pBBc (CTCF-70) restored the size of the protein made to 130 kDa. It is likely that
the introduced silent nucleotide substitutions altered particular stem-loop
formation in the CTCF mRNA so that the ribosome could pass freely.
The 70 kDa protein may be functionally important and there are indications which
make us believe that CTCF-70 does exist under normal physiological conditions. We showed previously
that approximately a quarter of total CTCF proteins in bursa was represented by
this form (
8
) and now we demonstrate that CTCF-70 can affect CTCF-130 transactivating capacity in COS6 cells (Fig.
6
). Since CTCF-70 does not have the DNA-binding and C-terminal domains, this apparent enhancement may result from
the competition of the two forms for a common transcriptional repressor(s).
CTCF-70 may have a similar role under normal physiological conditions (e.g. in
lymphocytes) providing post-translational tuning of CTCF-130 normal function(s) by competition with interacting protein
partners. Three proteins (p32, 80 and 97) have been specifically co-immuniprecipitated with CTCF-130 from COS6 cell extracts, however, one of them, p32, showed
interaction with the N-terminal domain (E. Klenova, unpublished observations). A two-hybrid system in yeast is currently being employed to identify and
clone other possible CTCF protein partners.
Previous investigations into
the existence of the pBBc mRNA under normal physiological conditions included
Northern blot analysis from various chicken cell lines and tissues. They
revealed two closely migrating bands of ~3.7 and 4.0 kb which are often impossible to resolve on agarose gels (
8
). It is tempting to speculate that the smaller message corresponds to the mRNA
with a part of its 5'-UTR missing (i.e., pBBc). Alternatively, this `artificial' mRNA may
serve as an appropriate substrate for the existent molecular apparatus
responsible for the similar function in the cell. It seems very intriguing that
a message with only 106 bp missing from the 5'-end produces a truncated protein. This may imply a translational
control mechanism of protein production carried out by the mRNA itself. In our
future studies we will attempt to fully elucidate the physiological
significance of this message.
We thank S. Mittnacht for the anti-GST antibody, C. Sutton for assistance in mass spectrometry, H. Paterson
for help with confocal microscopy. We are grateful to K. Davies and I. Craig
for helpful discussions. A. Robinson, I. Chernukhin and S. Christodoulou are
thanked for critical reading and suggestions on the manuscript. This work was
supported by the Human Frontiers Science Program Long-Term Fellowship and The Royal Society Research grant (E.M.K.), Special
Research Grant from The Research and Equipment Committee, University of Oxford
(R.E.L.), Cancer Research Compaign grants (G.H.G, R.H.N., S.U. and S.C.) and
Fogarty International Research Collaboration Award (V.V.L.).
REFERENCES
Return