Tài liệu Báo cáo khóa học: A multi-protein complex containing cold shock domain (Y-box) and polypyrimidine tract binding proteins forms on the vascular endothelial growth factor mRNA Potential role in mRNA stabilization pptx

A multi-protein complex containing cold shock domain (Y-box)
and polypyrimidine tract binding proteins forms on the
vascular
endothelial growth factor
mRNA
Potential role in mRNA stabilization
Leeanne S. Coles
1,
*, M. Antonetta Bartley
1,
*, Andrew Bert
1
, Julie Hunter
1
, Steven Polyak
2
, Peter Diamond
1
,
Mathew A. Vadas
1,3
and Gregory J. Goodall
1,3
1
Division of Human Immunology, The Hanson Institute, Institute of Medical and Veterinary Science;
2
Division of Biochemistry,
Department of Molecular Biosciences, The University of Adelaide;
3
Department of Medicine, The University of Adelaide,
North Terrace, Adelaide, South Australia, Australia

Vascular endothelial growth factor (VEGF) is a key regu-
lator of angiogenesis and post-transcriptional regulation
plays a major role in VEGF expression. Both the 5¢-and
3¢-UTR are required for VEGF post-transcriptional regu-
lation but factors binding to functional sequences within
the 5¢-UTR have not been fully characterized. We report
here the identification of complexes, binding to the VEGF
mRNA 5¢-and3¢-UTR, that contain cold shock domain
(CSD) and polypyrimidine tract binding (PTB) RNA
binding proteins. Analysis of the CSD/PTB binding sites
revealed a potential role in VEGF mRNA stability, in both
noninduced and induced conditions, demonstrating a gen-
eral stabilizing function. Such a stabilizing mechanism had
not been reported previously for the VEGF gene. We further
found that the CSD/PTB-containing complexes are large
multiprotein complexes that are most likely preformed in
solution and we demonstrate that PTB is associated with the
VEGF mRNA in vivo. Complex formation between CSD
proteins and PTB has not been reported previously. Analysis
of the CSD/PTB RNA binding sites revealed a novel CSD
protein RNA recognition site and also demonstrated that
CSD proteins may direct the binding of CSD/PTB com-
plexes. We found the same complexes binding to an RNA-
stabilizing element of another growth factor gene, suggesting
a broader functional role for the CSD/PTB complexes.
Finally, as the VEGF gene is also regulated at the tran-
scriptional level by CSD proteins, we propose a combined
transcriptional/post-transcriptional role for these proteins in
VEGF and other growth factor gene regulation.
Keywords: cold shock domain proteins; Y-box protein;

polypyrimidine tract binding protein; mRNA stabilization;
vascular endothelial growth factor.
VEGF is an essential regulator of angiogenesis that acts on
vascular endothelial cells to induce proliferation and
promote cell migration [1–3]. Disregulated VEGF expres-
sion is implicated in a number of diseases that are
characterized by abnormal angiogenesis [1–6]. In the case
of solid tumors, the overexpression of VEGF, produced in
response to activated oncogenes, growth factors or low
oxygen conditions (hypoxia), plays a major role in promo-
ting tumor angiogenesis and progression [1–3,7]. Both the
cancer cells themselves and nontumor support cells, such as
fibroblasts, are sources of VEGF [8]. In contrast, in the case
of coronary artery disease, inadequate VEGF expression
rather than VEGF overexpression, plays a role in disease
progression. A number of cell types, including cardiac
myocytes, fibroblasts and endothelial cells produce VEGF
in response to hypoxia, but this natural response is not
sufficient to prevent the further progression of heart disease
[9–11]. It is therefore important to understand the mecha-
nisms of VEGF regulation to develop means to control
VEGF expression.
Post-transcriptional regulation plays a major role in
VEGF expression, with regulation occurring at the level of
splicing, mRNA stability and translation [2,7]. The VEGF
mRNA is normally unstable and its stability is increased in
response to cytokines and stress conditions such as hypoxia
[7,11–14]. Regions in both the 5¢-and3¢-UTR have been
shown to be involved in VEGF mRNA stabilization
[7,12,13,15–18]. The presence of an internal ribosome entry

site (IRES) in the VEGF 5¢-UTR ensures continual
translation of the VEGF mRNA in stress conditions that
normally decrease cap-dependent translation [19–21]. Little
is known about the factors involved in VEGF post-
transcriptional regulation. Factors such as HuR and
hnRNPL have been implicated in hypoxic stability via their
Correspondence to L. S. Coles, Division of Human Immunology,
The Hanson Institute, Institute of Medical and Veterinary Science,
Frome Road., Adelaide, South Australia, 5000, Australia.
Fax: + 61 88 2324092, Tel.: + 61 88 2223432,
E-mail:
Abbreviations: CSD, cold shock domain; IRES, internal ribosome
entry site; VEGF, vascular endothelial growth factor.
*These authors contributed equally to this work.
(Received 16 October 2003, revised 14 December 2003,
accepted 16 December 2003)
Eur. J. Biochem. 271, 648–660 (2004) Ó FEBS 2004 doi:10.1111/j.1432-1033.2003.03968.x
actions on the VEGF 3¢-UTR [17,18] but factors involved in
stability or translational regulation have not been identified
on the 5¢-UTR.
The single-strand RNA and DNA binding, cold shock
domain (CSD) (also known as Y-box) proteins, play
diverse roles in both transcriptional and post-transcrip-
tional regulation of growth factor and stress response
genes [22–29]. CSD proteins have several family members
which are defined by the presence of a central highly
conserved 70 amino acid region called the cold shock
domain [24,25,29]. The central domain is required for
sequence-specific RNA binding, while the adjacent
C-terminal domain has a more nonspecific role in

stabilizing binding [24–27]. There are two types of
nongerm cell CSD proteins and these are called dbpB
(also known as YB-1, MSY-1, chkYB-1b, EF1A, p50 and
FRGY1) and dbpA (MSY4, chkYB-2 and YB2/RYBa).
DbpB and dbpA CSD proteins are ubiquitously expressed
and are highly conserved across species. Highly conserved
germ cell-specific CSD proteins also exist such as MSY-2
and FRGY2 [22–25,29]. In addition there are CSD-related
proteins such as UNR (upstream of N-ras) which contains
multiple conserved CSD domains [30]. CSD proteins
stabilize growth factor/stress response mRNAs in response
to stress signals [31–33] and also act as general mRNA
stabilizers [34–37]. In addition, CSD and CSD-related
proteins have been shown to play a role in cap-dependent
and [26,27,38–42] IRES-dependent [43–46] translation and
in RNA splicing [47,48]. In the case of the GM-CSF
(granulocyte-macrophage colony stimulating factor)
growth factor gene, CSD proteins have been shown to
play a combined role at both the transcriptional and post-
transcriptional levels [22,23,49,50]. As we have recently
shown a role for CSD proteins in regulation of the VEGF
gene at the transcriptional level [51], and given the diverse
functions of CSD proteins, relevant to VEGF expression,
we investigated a role for CSD proteins in VEGF post-
transcriptional regulation.
We now show here that CSD proteins can bind to both
the 5¢-and3¢-UTR of the VEGF mRNA. We find that
CSD proteins form a cytoplasmic complex on VEGF
mRNA that also contains the multifunctional single-
strand RNA/DNA binding protein, PTB [43–46,52–57],

and that the binding of this complex may be involved in
general VEGF mRNA stabilization. The CSD/PTB-con-
taining cytoplasmic complex also forms on a stability
element in the interleukin-2 (IL-2) 5¢-UTR suggesting a
similar mechanism of regulation of stability of growth
factor mRNAs.
Materials and methods
Plasmid constructs
The pGEM44, pGEM46 and pGEM47 constructs were
generated by cloning segments of the mouse VEGF
5¢-UTR, that were amplified by PCR from the pfVEGF
construct [15], into pGEM4Z (Promega). The pGEM44,
46 and 47 constructs contain, respectively, mouse VEGF
5¢-UTR sequences +1 to +325, +461 to +727 and +735 to
+1014 (relative to the transcription start site at +1) [58]
(Fig. 1). The pGEMV1 construct, containing the VEGF
5¢-UTR CSD site 1 sequences (+150 to +185) was
constructed by cloning double strand oligonucleotides (with
EcoRI 5¢-andHindIII 3¢-ends) into pGEM4Z. pGEMV37,
39, 15, 17 and 19 were similarly constructed, except that they
contained mutant versions of the CSD site 1 sequences
Fig. 1. The VEGF 5¢-UTR binds cytoplasmic and recombinant CSD
proteins. (A) Schematic of the mouse VEGF 5¢-UTR. The sequences
and coordinates (relative to the mRNA start site +1) [58] of consensus
CSD binding sites (CSD site 1,2) are indicated. The coordinates of
RNA probe sequences are also indicated. RNA probes were derived
from pGEM44, pGEM46 and pGEM47, respectively. (B) Balb/c 3T3
fibroblast cytoplasmic extracts were incubated without competitor (–)
or with unlabeled single-strand DNA competitor oligonucleotides
containing wild-type (CSDwt) and mutant (CSDmut) CSD binding

sites [49–51].
32
P-Labeled RNA probes (44 and 46) were then imme-
diately added and RNase T1 digested complexes analyzed by gel shift
assay. Cytoplasmic complexes CC44a, CC44b and CC46 and unbound
RNA probe are indicated. (C) Cytoplasmic extracts were preincubated
with anti-CSD polyclonal Ig (CSD), with preimmune serum (PI) or left
untreated (–), followed by addition of the labeled VEGF 44 RNA
probe in a gel shift assay. Increasing amounts of anti-CSD Ig were
added. Pre-immune sera was used at the maximal concentration used
for the anti-CSD Ig. Cytoplasmic complexes CC44a and CC44b are
indicated. (D) Recombinant GST-dbpB/YB-1 was incubated with
labeled 44, 46 and 47 RNA probes. Complexes were competed with
wild-type (CSDwt) or mutant (CSDmut) CSD binding site single-
strand DNA competitors or left untreated (–).
Ó FEBS 2004 CSD and PTB protein complexes on the VEGF mRNA (Eur. J. Biochem. 271) 649
(Figs 2 and 3). pGEMV25 and pGEMV27 were construc-
ted by cloning wild-type and mutant double strand oligo-
nucleotides containing the IL-2 5¢-UTR +1 to +35
sequences [31] (Fig. 4). pGEMVC1 was constructed by
cloning double strand oligonucleotides containing the
VEGF 3¢-UTR CSD site 3 (+1712 to +1747, relative to
the stop codon at +1, of the mouse VEGF 3¢-UTR) [59]
into pGEM4Z. pGEMVC2 and VC3 contained mutations
in the CSD site 3 sequence (Fig. 6).
The pfVEGF construct contained a reconstructed, tagged
VEGF cDNA sequence [15], composed of the entire VEGF
mouse 5¢-UTR (+1 to +1014), the coding region for the 164
aminoacidformofVEGFandtheVEGF3¢-UTR (+4 to
+2195). The major polyadenylation site is at +1861 [59]. A

Fig. 2. The VEGF 5¢-UTR CSD-containing cytoplasmic complexes also
contain PTB. (A) Schematic of the VEGF 5¢-UTR CSD site 1. The
coordinates for the VEGF RNA probes 44 and V1 are indicated rel-
ativetothestartoftheVEGF mRNA (+1) [58]. (B) Balb/c 3T3
fibroblast cytoplasmic extract was incubated with labeled VEGF 44 or
V1 RNA probes, followed by RNase T1 digestion in a gel shift assay.
The 44 and V1 RNA probes are derived from pGEM44 and
pGEMV1, respectively. Cytoplasmic complexes CC44a and CCV1
and unbound RNA probe are indicated. (C) Cytoplasmic complexes
CC44 and CCV1, in gel shift assay gels, were exposed to UV light to
cross-link proteins in each complex to RNA. Cross-linked proteins
were then analyzed by SDS/PAGE and the sizes of cross-linked pro-
teins were calculated by subtraction of the molecular weight of bound
RNA probe. The sizes of cross-linked proteins are indicated. Cross-
link analysis is not quantitative as different proteins will crosslink to
different extents. (D) Cytoplasmic extracts were incubated with unlabe-
led wild-type or mutant CSD (CSDwt, mut) or PTB (PTB wt, mut) [52]
binding site single-strand DNA oligonucleotides, or left untreated (–).
Labeled V1 RNA probe was then immediately added and complexes
analyzed in a gel shift assay. The CCV1 complex is indicated. (E)
Cytoplasmic extracts were preincubated with an anti-PTB monoclonal
antibody (PTB), with a control anti-GM-CSF monoclonal antibody
(GM), or without antibody (–). Labeled V1 RNA probe was then
added and complexes analyzed in a gel shift assay.
Fig. 3. Sequence requirement for VEGF 5¢-UTR CCV1 CSD/PTB
complex formation-PTB complexes form on the VEGF mRNA in vivo.
(A) The sequence of the VEGF 5¢-UTR V1 RNA probe is shown and
consensus CSD and PTB protein binding site sequences are indicated.
The 3¢ consensus PTB site is also found, in this report, to bind
recombinant CSD protein (labeled CSD*). The sequences of mutant

RNA probes are given under the V1 sequence. Only those bases that
are changed in the mutant probes are indicated. The RNA probes were
generated from pGEMV1, V37, V39, V15, V17 and V19 constructs,
respectively. (B) Balb/c 3T3 fibroblast cytoplasmic extracts were
incubated with labeled wild-type (V1) and mutant VEGF 5¢-UTR CSD
site 1 RNA probes in a gel shift assay. The CCV1 cytoplasmic complex
is indicated. (C) Recombinant GST-dbpB/YB-1 and GST-PTB were
incubated with labeled wild-type (V1) and mutant (V37, V39) VEGF
5¢-UTR RNA probes in a gel shift assay. The recombinant protein
complexes are indicated. (D) PTB binding to VEGF mRNA in vivo was
investigated using an RNA immunoprecipitation assay. Cytoplasmic
RNA/protein complexes (prepared in the presence of RNase inhibi-
tors) were immunoprecipitated with anti-PTB monoclonal Ig (PTB),
with an IgG2 isotype control (control), or without antibody (–). VEGF
mRNA in RNA extracted from immunoprecipitated complexes was
detected by RT-PCR. The VEGF PCR product is indicated.
650 L. S. Coles et al. (Eur. J. Biochem. 271) Ó FEBS 2004
polylinker is positioned between the coding region and the
3¢-UTR sequences to distinguish pfVEGF mRNA from
endogenous VEGF mRNA. pfVEGFdel contains deletions
of the site 1, 2 and 3 CSD sites (Fig. 7). The sequences +156
to +179 (CSD site 1) and +650 to +666 (CSD site 2) of the
5¢-UTR were deleted and the sequences +1727 to +1740
(CSD site 3) of the 3¢-UTR were deleted.
Construction of the expression vector producing recom-
binant GST-dbpB/YB-1 (pGEXBT) has been described
previously [49,51]. pGEXPTB, for production of recombin-
ant GST-PTB, was constructed by cloning of a 1.6kb EcoRI
fragment from pcDNA3PTB (gift from T. Cooper, Baylor
College of Medicine, Houston, TX, USA), coding for

human PTB, into pGEX4T-2.
Oligonucleotides
Oligonucleotides for cloning into pGEM4Z and for use as
competitors in gel shift assays were synthesized by Gene-
works (Adelaide, Australia) and purified from nondenatur-
ing polyacrylamide gels. Single-strand oligonucleotides for
competition of CSD protein-containing complexes were
from the human granulocyte-macrophage-colony stimula-
ting factor (GM-CSF)gene.Thewild-type(CSDwt)and
mutant sequences (CSD site mutant; CSDmut) have been
described previously (GM- and GMm23-, respectively)
[49–51]. The CSD wild-type sequence binds both dbpA and
dbpB CSD proteins. The wild-type (PTBwt) and mutant
PTB (PTBmut) competitor single-strand DNA oligonucleo-
tides are from the transferrin gene (DR1 sense and DR1
sense mut1, respectively) [52].
RNA probe preparation
32
P-labeled RNA probes for gel shift analysis or RNase
protection assays were generated by in vitro transcription
from linearized plasmid templates (pGEM4Z constructs)
using SP6 (for pGEM44,46,47) or T7 (for pGEMV1, V25
and VC1) RNA polymerase (Promega) and [
32
P]UTP[aP].
Probes for RNase protection assays were processed as
previously described [15]. Probes for gel shift assays were
purified from nondenaturing polyacrylamide gels and eluted
into RNase-free water at 56 °C.
Preparation of recombinant and cytoplasmic proteins

The Escherichia coli strain MC1061 transformed with
pGEXBT or pGEXPTB was induced with isopropyl thio-
b-
D
-galactoside to produce recombinant GST-dbpB/YB-1
and GST-PTB [49,51]. The fusion proteins for gel shift
analysis were purified on glutathione-Sepharose beads
(Promega). Cytoplasmic extracts were produced according
to the method of Schrieber et al. [60].
FPLC gel filtration of cytoplasmic extracts
Cytoplasmic extract from Balb/c 3T3 fibroblasts was
applied at a flow rate of 0.35 mLÆmin
)1
to a Superdex 200
column (10 mm diameter, 20 mL bed volume) pre-equili-
brated with buffer containing 150 m
M
KCl, 20 m
M
Tris/
HCl pH 7.6, 20% glycerol, 1.5 m
M
MgCl
2
,2m
M
dithio-
threitol, 0.4 m
M
phenylmethanesulfonyl fluoride and 1 m

M
Na
3
VO
4
. The CCV1 complex was eluted with the same
buffer and 0.5 mL fractions collected. The molecular mass
of the complex was estimated from the column by
comparison with the elution volumes of c-globulin, bovine
serum albumin, ovalbumin, myoglobin and vitamin B12.
Antibodies
The anti-CSD antibody is a rabbit polyclonal Ig raised
against a peptide conserved in dbpA and dbpB/YB-1 CSD
proteins across species [49,51]. The anti-PTB Ig is a mouse
monoclonal antibody (BB7; gift from D. Black, UCLA,
Los Angeles, CA, USA). A mouse monoclonal anti-
(GM-CSF) Ig (gift from A. Lopez, Hanson Institute,
IMVS, Adelaide, Australia) and an IgG2 monoclonal
Fig. 4. CCV1 complex formation on the IL-2 5¢-UTR stability element
in fibroblasts and Jurkat T cells. (A) Sequence of the IL-2 5¢-UTR wild-
type probe V25 with consensus CSD and PTB sites indicated. The
region +1 to +22 is involved in IL-2 mRNA stabilization in T cells
[31]. The V27 mutant sequence is shown with only those bases differing
from the wild-type sequence shown. (B) Balb/c 3T3 fibroblast cyto-
plasmic extracts were incubated with labeled VEGF (V1) and IL-2
(V25,V27) 5¢-UTR RNA probes, and analyzed by gel shift. The CCV1
cytoplasmic complex is indicated. (C) The Balb/c 3T3 CCV1 com-
plexes binding to the VEGF (V1) and IL-2 (V25) RNA probes (after
RNase T1 digestion) were exposed to UV light, in gel shift gels, to
cross-link proteins to RNA. Cross-linked proteins were sized by SDS/

PAGE. The sizes of cross-linked proteins were calculated by subtrac-
tion of molecular masses of bound RNA probes. (D) Balb/c 3T3
fibroblast and Jurkat T cell cytoplasmic extracts were incubated with
labeled VEGF (V1) and IL-2 (V25) RNA probes, digested with RNase
T1 and analyzed in a gel shift assay. The CCV1 cytoplasmic complex is
indicated. (E) The Jurkat T cell CCV1 cytoplasmic complex binding to
the IL-2 V25 RNA probe was analyzed by UV cross-link analysis as
described above. The sizes of cross-linked proteins are indicated.
Ó FEBS 2004 CSD and PTB protein complexes on the VEGF mRNA (Eur. J. Biochem. 271) 651
antibody isotype control (Silensus, Boronia, Victoria, Aus-
tralia) were used as controls for the anti-PTB antibody in gel
shift assays and RNA immunoprecipitations, respectively.
RNA gel shift analysis, competitions and antibody
analysis
RNA gel shifts were performed using
32
P-labeled RNA
probes in a 10 lL reaction mix of 0.5· TM buffer [49–51]
containing 200 m
M
KCl, 1 lg poly(dI.dC), 100 ng tRNA,
1 lg bovine serum albumin and either 1 lg cytoplasmic
extract or 25 ng recombinant protein (GST-dbpB or GST-
PTB). Reactions were incubated at 4 °C for 20 min,
followed by treatment with or without RNase T1 (Worth-
ington Biochemical Corp., NJ, USA) and analyzed on
6% nondenaturing polyacrylamide gels. Competition with
single-strand DNA oligonucleotides was performed by
addition of protein and 50 ng of unlabeled probe, followed
by immediate addition of the

32
P-labeled RNA probe.
Antibody blocking experiments were performed by incuba-
ting protein and antibody for 5 min before adding the
32
P-labeled probe. Antibodies did not degrade RNA probes
under the gel shift conditions used.
UV cross-linking
Cytoplasmic extracts were bound to
32
P-labeled RNA
probes in a 25 lLgelshiftreactionandfractionatedona
6% polyacrylamide gel as described above. The gel was
exposed to UV light (340 nm) for 15 min and retarded
complexes were excised after exposure overnight to X-ray
film. Protein in excised bands was analyzed by 12% SDS/
PAGE as described previously [49–51].
RNA immunoprecipitation assay
Balb/c 3T3 fibroblast extracts were prepared, as described
above, in the presence of RNase inhibitors (Promega) and
incubated with or without anti-PTB monoclonal Ig or with
an IgG2 isotype control for 60 min. RNase inhibitors were
required to prevent the loss of RNA from extracts. Protein
A sepharose CL-4B (Pharmacia, Biosciences, Uppsala,
Sweden) was added and further incubated for 60 min.
Sepharose was extracted for bound RNA (TRIzolÒ
reagent, Invitrogen). RNA was reverse transcribed using
Superscript II (Promega) and a PCR assay for VEGF
cDNA was performed using oligonucleotides from the
mouse VEGF cDNA of 5¢-CACAGACTCGCGTTGCA-3¢

and 5¢-TGGGTGGGTGTGTCTAC-3¢. PCR products
were analyzed by agarose gel electrophoresis. The VEGF
PCR product is approximately 400 bp.
Cell culture, stable transfection and cell stimulation
Mouse Balb/c 3T3 fibroblasts and rat C6 glioma cells were
grown in Dulbecco’s modified Eagle’s medium with 10%
fetal bovine serum. Jurkat T cells were cultured in RPMI
media with 10% fetal bovine serum. For cytoplasmic
extracts, cells were grown in normoxic conditions (normal
oxygen; 20% O
2
). For the production of stably transfected
cell lines, C6 glioma cells were transfected with linearized
pfVEGF or pfVEGfdel plasmids using lipofectamine
TM
2000 (Gibco BRL Life Technologies, Melbourne, Australia)
according to the manufacturer’s directions. Cells were
grown for 24–48 h and selected in 400 lgÆmL
)1
G418 [15].
Serum stimulation of stably transfected cell lines was as
described previously [15]. Hypoxic conditions (1% O
2
)were
generated in a hypoxic chamber (Edwards Instrument
Company, Sydney, Australia).
Analysis of mRNA stability
in vivo
Stable transfectants (pfVEGF or pfVEGFdel) were serum
stimulated (time 0) and concurrently incubated under

normoxic or hypoxic conditions for 1, 1.5, 2, 3 or 4 h.
Serum stimulation provides a brief pulse of transcription
from the c-fos promoter in pfVEGF/pfVEGFdel con-
structs, allowing subsequent degradation of the mRNA to
be monitored as previously described by us in analysis of the
pfVEGF construct [15]. This system allows determination
of mRNA stability directly, rather than using indirect means
such as nonspecific inhibitors of transcription. RNA was
isolated from treated cells using TRIzol
r
reagent (Invitro-
gen) according to the manufacturers instructions, and
pfVEGF/pfVEGFdel mRNA was detected by RNase pro-
tection analysis using a
32
P-labeled transcript covering the
polylinker sequence in the pfVEGF/pfVEGFdel constructs
as previously described [15]. Neomycin phosphotransferase
(neo) mRNA expressed from pfVEGF/pfVEGFdel con-
structs was detected as described [15]. Protected RNAs were
separated on denaturing polyacrylamide gels and the
amounts of specific
32
P-labeled protected pfVEGF/pfVEGF-
del or neo mRNAs were quantitated by PhosphoImager
analysis (Molecular Dynamics, Sunnyvale, CA, USA).
Levels of pfVEGF/pfVEGFdel mRNA (with time 0 levels
subtracted) were normalized with respect to the levels of neo
mRNA at each time point.
Results

The VEGF 5¢-UTR binds cytoplasmic and recombinant
CSD proteins
Sequence specific RNA binding sites for CSD proteins
have been determined in a few genes but a consensus
sequence has not been established. Analysis of the prota-
mine 1 (Prm1) 3¢-UTR has revealed a preferred binding site
of 5¢-U/C/A–C/A–C–A–U/C–C–A/C/U-3¢ for mouse CSD
proteins [38–40]. This sequence is consistent with a prefer-
red sequence for Xenopus CSD proteins (FRGY1/2) of
5¢-AACAUCU-3¢ [61] and with a 5¢-ACCACC-3¢ sequence
from the Rous Sarcoma virus LTR that binds chicken CSD
proteins [41].
Given a potential role for CSD proteins in VEGF post-
transcriptional regulation, the VEGF 5¢-UTR was exam-
ined for CSD protein binding sites. Two potential sites at
+157 and +650 were observed. These were named CSD site
1 and CSD site 2 and have sequences of 5¢-AACCU
CU-3¢ and 5¢-AACUUCU-3¢, respectively (Fig. 1A). No other
potential CSD protein binding sequences were observed.
To determine if the VEGF 5¢-UTR could bind cytoplas-
mic CSD complexes,
32
P-labeled RNA probes 44 (+1 to
+325) and 46 (+461 to +727) covering the potential CSD
sites (Fig. 1A) were bound to cytoplasmic extracts from
652 L. S. Coles et al. (Eur. J. Biochem. 271) Ó FEBS 2004
Balb/c 3T3 fibroblasts and analyzed by gel shift assay
(Fig. 1B). The 44 and 46 probes formed strong complexes
with cytoplasmic proteins (CC44a, b and CC46, respect-
ively) and these complexes were readily competed by a

single-strand DNA oligonucleotide (CSDwt) from the GM-
CSF gene that is known to bind CSD proteins [49,50]. These
complexes were less readily competed by a mutant version
of the GM-CSF CSD oligonucleotide (CSDmut), suggest-
ing the presence of CSD proteins (Fig. 1B). In support of
this, formation of complexes on the 44 RNA probe were
blocked by preincubation of extracts with increasing
amounts of an anti-CSD polyclonal Ig (CSD), whereas
preimmune serum (PI) had no effect (Fig. 1C). Cytoplasmic
complexes containing CSD proteins can therefore bind the
VEGF 5¢-UTR.
To further support a role for CSD proteins binding the
VEGF 5¢-UTR, it was observed that a recombinant CSD
protein GST-dbpB/YB-1 could bind to both the 44 and 46
RNA probes but not to a probe which does not contain
a potential CSD site (probe 47; +735 to +1014). The 47
probe contains sequences required for mouse VEGF IRES
activity [19,20]. As for cytoplasmic complexes, GST-dbpB/
YB-1 binding was specifically competed by the CSD wt
oligonucleotide (Fig. 1D). Both cytoplasmic and recombin-
ant CSD protein complexes are therefore forming on the
VEGF 5¢-UTR.
The VEGF 5¢-UTR CSD-containing cytoplasmic complexes
also contain PTB
To localize the binding site for the major complex on the 44
RNA probe (CC44a), Balb/c 3T3 fibroblast cytoplasmic
extract was bound, in a gel shift assay, to a shorter
32
P-labeled RNA probe, containing the proposed CSD
binding site 1 (V1; +150 to +185, Fig. 2A). A single major

complex (CCV1) formed on the V1 probe and migrated in a
similar position to the CC44a complex on the 44 RNA
probe (Fig. 2B). To verify that the CC44a and CCV1
complexes were the same, complexes were analyzed by UV
cross-linking (Fig. 2C). The CC44a and CCV1 complexes,
that had been separated in a gel shift gel, were exposed
to UV light to cross-link proteins in complexes to their
respective RNA probes. Cross-linked proteins were then
separated by SDS/PAGE and the sizes of cross-linked
proteins were calculated by subtraction of molecular masses
of bound RNA fragments. The number and size of proteins
cross-linked to RNA was identical for the 44 and V1 RNA
probes. We similarly found that the 46 RNA probe binding
complex (CC46), that contains the CSD site 2, gave an
identical cross-link pattern (data not shown).
As expected the CSD-containing cytoplasmic complexes
binding to the CSD site 1 (and CSD site 2) contained a
protein of 50 kDa, consistent with the size of the CSD
protein, dbpB (also known as YB-1/p50) [42]. Additional
proteins in the complex had sizes of 60, 27 and 12 kDa. The
single CSD-containing cytoplasmic RNA/protein complex
therefore contains at least four different proteins. It has been
reported that PTB, another single-strand RNA/DNA
binding protein [43–46,52–57] can bind to a 50 base region
spanning the CSD site 1 sequence in the human VEGF
5¢-UTR [21]. Given that we have detected a 60 kDa protein
of the approximate size for PTB (57 kDa), we further
investigated the CSD site 1, CCV1 complex. The presence of
PTB protein in the CCV1 complex was confirmed in a gel
shift assay, by competition of the CCV1 complex with an

unlabeled single-strand DNA oligonucleotide probe from
the transferrin gene (PTB wt), that binds PTB [52]. The
CCV1 complex was not readily competed by a transferrin
gene oligonucleotide with a mutant PTB site (PTB mut)
(Fig. 2D). We confirmed using recombinant proteins that
the PTB wt oligonucleotide could not bind CSD proteins
and hence was specific for PTB (data not shown). The
presence of PTB was further confirmed by preincubation of
cytoplasmic extract with an anti-PTB monoclonal Ig (PTB)
before binding to V1 RNA probe in a gel shift assay. The
formation of the CCV1 complex was blocked by the anti-
PTB Ig (PTB) but not by an irrelevant monoclonal antibody
(anti-GMCSF; GM) (Fig. 2E).
Taken together, this data demonstrates that the CSD-
containing cytoplasmic complex forming on the VEGF
5¢-UTR contains PTB in addition to further unknown
proteins of 27 and 12 kDa. The ability of anti-CSD and
PTB antibodies and DNA competitors to effectively
compete complex formation demonstrates the dependence
on the presence of both CSD and PTB proteins to form the
CCV1 complex.
VEGF 5¢-UTR CCV1 complex formation requires the
consensus 5¢-ACCUCUU-3¢ sequence and a downstream
5¢-UUUUCUU-3¢ sequence
To determine the sequences required for CCV1 complex
formation on the VEGF 5¢-UTR CSD site 1 RNA probe
(V1), Balb/c 3T3 fibroblast cytoplasmic extract was bound
to mutant versions of the V1 probe (Fig. 3A) and analyzed
in a gel shift assay (Fig. 3B). Mutations were made in the
predicted CSD protein binding site, 5¢-ACCUCUU-3¢,and

also in an adjacent sequence containing a potential PTB
site 5¢-UUUUCUU-3¢.5¢-UCUU-3¢ sequences flanked by
pyrimidine residues are commonly found in PTB binding
sites [52–55,57]. Mutation of either sequence, by a block
mutation (V37, V39) or by mutation of the central UC
residues to AA (V15, V17) reduced CCV1 binding,
suggesting a role for both sequences in complex formation.
Consistent with this, a double mutation (V19) abolished
CCV1 complex formation (Fig. 3B).
To investigate the individual roles CSD and PTB proteins
may play in directing CCV1 complex binding to the
5¢-ACCUCUU-3¢ and 5¢-UUUUCUU-3¢ sequences,
recombinant CSD (GST-dbpB/YB-1) and PTB (GST-
PTB) binding to wild-type (V1) and mutant (V37, V39)
RNA probes was examined (Fig. 3C). Consistent with
CCV1, GST-dbpB/YB-1 binding was reduced by mutation
of both sites (V37, V39) whereas PTB binding was only
slightly reduced by mutation of the 3¢-pyrimidine-rich
sequence (V39). The presence of the adjacent poly U stretch
may provide an alternative contact site for the recombinant
PTB protein. CSD proteins may play a larger role in
directing CCV1 complex formation than PTB via their
ability to bind both the predicted CSD site, 5¢-ACCUC
UU-3¢, and the downstream 5¢-UUUUCUU-3¢ sequence
required for CCV1 complex formation.
To confirm that PTB-containing complexes can form on
the VEGF mRNA in vivo an RNA immunoprecipitation
Ó FEBS 2004 CSD and PTB protein complexes on the VEGF mRNA (Eur. J. Biochem. 271) 653
assay was performed. Cytoplasmic RNA/protein complexes
from Balb/c 3T3 fibroblasts were immunoprecipitated with

an anti-PTB monoclonal antibody and RNA extracted
from immunoprecipitated complexes was assayed by
RT-PCR for mouse VEGF mRNA sequences (Fig. 3D).
Cytoplasmic extracts, for immunoprecipitation, were made
in the presence of RNase inhibitors to prevent RNA loss.
VEGF mRNA was readily detected by RT-PCR in samples
immunoprecipitated with anti-PTB monoclonal Ig (PTB).
VEGF mRNA was not, however, detected in immunopreci-
pitations performed with an IgG2 monoclonal antibody
isotype control or without the addition of antibody (–).
An IL-2 5¢-UTR stability element binds the same
cytoplasmic complex as the VEGF 5¢-UTR
The IL-2 5¢-UTR contains sequences, at +1 to +22, that are
required for mRNA stabilization in T cells. Both dbpB/YB-
1 and another RNA binding protein, nucleolin, bind to this
region and are involved in stabilization [31]. Inspection of
the IL-2 5¢-UTR stability element revealed a sequence,
5¢-ACUCUCUU-3¢,at+4to+11,thatwasverysimilarto
the VEGF CSD site 1 CSD consensus sequence (Fig. 4A).
The ability of the IL-2 sequence to bind the CCV1 complex
was tested in a gel shift assay using Balb/c 3T3 fibroblast
cytoplasmic extract (Fig. 4B). An RNA probe containing
the +1 to +35 IL-2 5¢-UTR sequences (V25) bound a
similarly migrating complex to that observed on the VEGF
V1 probe (CCV1), and this complex was abolished by
mutation of the 5¢-ACUCUCUU-3¢ sequence (V27). The
V27 block mutation is reported to reduce IL-2 mRNA
stability in T cells [31]. UV cross-link analysis of the IL-2
complex revealed that it was identical to the VEGF 5¢-UTR
CCV1 complex (Fig. 4C). An identical complex from

fibroblast extracts can therefore form on both the VEGF
and IL-2 genes.
For the CCV1 complex to be of relevance to the
regulation of expression of the IL-2 gene, it was important
to determine if the complex could be formed using T cell
extracts. Binding of the IL-2 V25 probe to Jurkat T cell
cytoplasmic extracts revealed the formation of a complex
comigrating with the fibroblast CCV1 complex (Fig. 4D).
UV cross-linking demonstrated that the fibroblast and
T-cell complexes were identical (Fig. 4E). The CCV1
complex therefore forms on a functional element in the
IL-2 5¢-UTR in T cells.
The VEGF 5¢-UTR CCV1 complex may be preformed
To determine if cytoplasmic CSD/PTB-containing VEGF
5¢-UTR complexes can form in the absence of RNA, Balb/c
3T3 fibroblast extract was fractionated by FPLC gel
filtration and the fractions incubated with wild-type (V1)
or mutant (V19; Fig. 3) VEGF CSD site 1 RNA probes in a
gel shift assay (Fig. 5). CCV1 complex formation, binding
to the V1 probe, was observed in fractions with an
approximate molecular mass range of 400–490 kDa (frac-
tions 6, 7). No other complexes were observed across the
range of fractions analyzed (from 10 to 1000 kDa) (data not
shown). CCV1 complex formation, in fractions 6 and 7, was
abolished by the V19 mutation, verifying the nature of these
complexes (Fig. 5). These data indicate that the CCV1
complex is preformed in solution. The presence of a higher
order preformed complex suggests the possibility that CSD
and PTB may interact. Consistent with this, CSD proteins
have been shown to interact with a number of partner

proteins in solution [22,23].
Involvement of the VEGF 3¢-UTR in binding CSD
and PTB proteins
As sequences in the 3¢-UTR are also involved in post-
transcriptional regulation of VEGF expression, the mouse
VEGF 3¢-UTR [59] was examined for potential CSD
protein binding sites. A single site at +1727, relative to the
stop codon at +1, was observed with a sequence of
5¢-AACAUCA-3¢. This sequence is an exact match to the
preferred binding site for mouse CSD proteins [38,40] and,
as for the VEGF 5¢-UTR sites, has a potential PTB site,
5¢-UCUU-3¢, immediately downstream at +1736 (Fig. 6A).
We used gel shift assays to examine whether a CSD/PTB
complex binds to this region in the 3¢-UTR. An RNA probe
(VC1) containing sequences +1712 to +1747 of the mouse
VEGF 3¢-UTR was bound to Balb/c 3T3 fibroblast
cytoplasmic extracts and two major complexes were
observed. The faster migrating complex (CCVC1) was
competed more readily with wild-type (wt) than mutant
(mut) PTB and CSD protein binding oligonucleotides,
suggesting that this complex contains PTB and CSD
proteins (Fig. 6B). As expected, mutation of either the
potential CSD site (VC2) or the PTB site (VC3) reduced the
formation of the CCVC1 complex (Fig. 6C). Consistent
with this, recombinant GST-PTB and GST-dbpB/YB-1
bound to the VC1 probe and PTB and CSD protein binding
were reduced by mutations in the PTB (VC3) and CSD
protein (VC2) binding sites, respectively. As was observed
for the 5¢-UTR, CSD binding was also reduced by mutation
of the potential PTB site (VC3) (Fig. 6D).

Fig. 5. Gel filtration fractionation of the VEGF 5¢-UTR CCV1 com-
plex. Balb/c 3T3 fibroblast cytoplasmic extract was fractionated by
FPLC gel filtration and fractions assayed by incubation with labeled
VEGF 5¢-UTR wild-type (V1) and mutant (V19) RNA probes in a gel
shift assay. Consecutive 0.5 mL fractions containing CCV1 binding
activity are shown. The approximate sizes of protein fractions, deter-
mined by comparison to elution profiles of protein standards, is given
in kDa. The CCV1 complex is indicated.
654 L. S. Coles et al. (Eur. J. Biochem. 271) Ó FEBS 2004
Hence as for the 5¢-UTR, both the potential CSD and
PTB 3¢-UTR sites are required for cytoplasmic complex
(CCVC1) formation, with recombinant PTB primarily
contacting the downstream PTB site and CSD protein
contacting both sites.
The
VEGF
mRNA CSD/PTB binding sites play a role
in
VEGF
mRNA stability
The binding of common cytoplasmic complexes to the IL-2
5¢-UTR stability element and the VEGF CSD/PTB sites,
suggested a possible role for these sites in VEGF mRNA
stability. CSD proteins have been shown to be involved in
both inducible mRNA stabilization, as is observed for the
IL-2 mRNA [31–33], and general mRNA stabilization
[34–37]. PTB proteins may also play a role in general
mRNA stabilization [62].
To investigate a role for the VEGF mRNA CSD/PTB
sites, we analyzed the in vivo stability of mRNAs produced

from constructs containing tagged VEGF mRNA coding
Fig. 6. CSD and PTB proteins bind to the VEGF 3¢-UTR. (A)
Sequence of the VEGF 3¢-UTR CSD site 3 RNA probe (VC1). The
sequences represent +1712 to +1747 of the mouse VEGF 3¢-UTR,
relative to the stop codon at +1 [59]. Consensus CSD and PTB binding
sequences are indicated. CSD* indicates that recombinant CSD pro-
tein can also contact the PTB site. Mutant RNA probe sequences
(VC2, VC3) are shown with only those bases that differ from the wild
type indicated. RNA probes were generated from pGEMVC1, VC2
and VC3 constructs. (B) Balb/c 3T3 fibroblast cytoplasmic extract was
incubated with wild-type and mutant CSD (CSDwt,mut) or PTB
(PTBwt,mut) binding site single-strand DNA competitors or left
untreated (–). Labeled VC1 probe was then immediately added and
complexes analyzed in a gel shift assay. The CCVC1 complex and
unbound RNA probe are indicated. (C) Cytoplasmic extracts were
incubated with VEGF 3¢-UTR wild-type (VC1) and mutant (VC2,
VC3) RNA probes and analyzed in a gel shift assay. (D) Recombinant
GST-dbpB/YB-1 and GST-PTB were incubated with VEGF 3¢-UTR
wild-type (VC1) and mutant (VC2,VC3) RNA probes. Recombinant
complexes are indicated.
Fig. 7. Deletion of the VEGF CSD/PTB sites affects VEGF mRNA
stability in normoxic and hypoxic conditions. (A) Diagrammatic rep-
resentation of the pfVEGF [15] and pfVEGFdel constructs. The c-fos
promoter and sequences encoding the mouse VEGF mRNA are
indicated. A poly linker is located between the protein coding and
3¢-UTR sequences as a tag for detection of pfVEGF and pfVEGFdel
mRNA. The sequences deleted from the CSD sites 1 (+156 to +179),
2 (+650 to +666) and 3 (+1727 to +1740) in the pfVEGFdel con-
struct are indicated with dashes. (B) Stable transfectants, containing
pfVEGF or pfVEGFdel were serum stimulated at time 0 to induce a

brief pulse of RNA expression from the c-fos promoter from which
mRNA degradation can be followed [15]. Cells were simultaneously
treated (at time 0) under normoxic or hypoxic conditions and the levels
of pfVEGF/pfVEGfdel mRNA expressed from constructs determined
by RNase protection assay. mRNA levels were normalized with
respect to neo mRNA. Serum stimulation increased transfected
mRNA levels approximately 10-fold in both the pfVEGF and
pfVEGFdel stable transfectants and was maximal at the 1 h time
point. The levels of pfVEGF and pfVEGFdel mRNAs were approxi-
mately 1.7-fold the levels of endogenous VEGF mRNA in respective
cell lines at this time (data not shown). The percentage mRNA
remaining, relative to the mRNA levels at the 1 h time point (given as
100%), is shown as a linear plot for experiments performed under
normoxic and hypoxic conditions. Data is the average of five experi-
ments. RNase protection gel data is shown below the linear plot for
one representative experiment. Three repeats were performed for each
time point. The pfVEGF, pfVEGFdel and neo transcripts are indi-
cated. (C) Presentation of data in (B) as a log plot.
Ó FEBS 2004 CSD and PTB protein complexes on the VEGF mRNA (Eur. J. Biochem. 271) 655
sequences with either wild-type or deleted CSD/PTB sites
(pfVEGF and pfVEGFdel, respectively) (Fig. 7A). We have
reported previously the use of pfVEGF in stabilization
experiments [15]. pfVEGF or pfVEGFdel stable transfect-
ants were serum stimulated and simultaneously exposed to
hypoxic or normoxic conditions. Transfected wild-type or
mutant VEGF mRNA levels were then assayed by RNase
protection assay at time intervals following serum stimula-
tion to determine the stability of respective mRNAs
(Fig. 7B,C).
It can be seen in Fig. 7(B,C) that degradation of the wild-

type VEGF mRNA (pfVEGF) occurs less rapidly than for
the mutant mRNA (pfVEGF del) under both normoxic and
hypoxic conditions, indicating that the CSD/PTB sites play
a role in stabilizing the VEGF mRNA in both noninduced
and hypoxia-induced conditions. The wild-type pfVEGF
mRNA is 1.3-fold more stable than the CSD/PTB site
deleted mRNA. Mutation of the CSD/PTB sites, however,
had no effect on the ability of hypoxia to increase mRNA
stability relative to that seen under normoxic conditions.
Both wild-type and mutant VEGF mRNAs were stabilized
approximately 1.4-fold by hypoxia. The VEGF CSD/PTB
sites therefore are not involved in induced stabilization in
response to hypoxia but appear to be involved in general
stabilization of the VEGF mRNA. Interestingly the pres-
ence of the CSD/PTB sites confers a similar degree of
increased stability to the VEGF mRNA to that produced
under hypoxic conditions.
Discussion
Inappropriate or inadequate expression of VEGF plays a
key role in the progression of a number of diseases [1–6]. It is
therefore important to determine the processes involved in
regulation of VEGF expression. We had previously shown
that cold shock domain (CSD) (or Y-box) proteins regu-
lated VEGF expression at the transcriptional level in the
nucleus [51]. We show here that CSD proteins may also play
a role in post-transcriptional regulation of VEGF expres-
sion in the cytoplasm, in conjunction with another single-
strand RNA/DNA binding protein, PTB.
Conserved CSD/PTB binding sites in the
VEGF

mRNA
5¢- and 3¢-UTR
The 5¢-and3¢-UTR of the VEGF mRNA are involved in
post-transcriptional regulation [7,11–15,17,17–21] and we
have identified CSD/PTB protein binding sites in both
these regions. Two sites were found in the 5¢-UTR (CSD
sites 1 and 2) and one site was found in the 3¢-UTR (CSD
site 3) (Fig. 8A). All three sites contain a sequence that is
similar to a preferred RNA binding sequence determined
for the mouse CSD proteins MSY1, 2 and 4 [38–40]. This
preferred sequence is consistent with the RNA binding
sites identified for chicken, frog and human CSD proteins
[31,41,47,61] (Fig. 8B). The VEGF sequences all show a
substitution of the fourth position of the preferred mouse
sequence from an A to a C or U residue. Potential PTB
binding sequences, 5¢-UCUU-3¢ (flanked by U/C residues)
[52–55,57], were located immediately downstream of the
consensus CSD sequences in both the 5¢-and3¢-UTR
CSD/PTB sites. An additional potential PTB site over-
lapped the consensus CSD sequence in the VEGF 5¢-UTR
site 1 (Fig. 8A).
Consistent with the presence of potential PTB binding
sites, cytoplasmic complexes binding to the VEGF 5¢-and
3¢-UTR sites contained both CSD and PTB proteins and it
was observed that both the consensus CSD sequence and
the downstream PTB sequence, within these sites, were
required for full complex formation. The ability of both
antibody and oligonucleotide competitors to reduce or
abolish complex formation demonstrated that CSD and
PTB proteins were simultaneously bound to VEGF RNA.

The ability of CSD and PTB proteins to bind to the 5¢-and
3¢-UTR sequences was further demonstrated by the binding
of recombinant PTB and dbpB/YB-1 CSD protein to these
sites. Importantly, we found that PTB-containing com-
plexes could be detected on the VEGF mRNA in vivo.PTB
binding was primarily affected by the mutation of the
downstream PTB consensus sequences, while surprisingly,
recombinant CSD binding was affected by mutation of
either the consensus CSD site or the downstream PTB
sequence. CSD proteins therefore recognize not only the
expected consensus RNA sequence but also the VEGF PTB
binding sequences. The binding of CSD proteins to PTB
Fig. 8. Comparison of CSD protein RNA binding sites. (A) The
sequences of the VEGF 5¢-and3¢-UTR CSD site 1, 2 and 3 sequences
are shown and consensus CSD (5¢)andPTB(3¢) protein binding sites
are underlined. Both sequences are required for cytoplasmic CSD/PTB
complex formation and recombinant CSD protein binding. (B) A
comparison of CSD protein RNA binding sites is shown relative to a
preferred sequence derived for the mouse MSY1/2/4 CSD proteins
[38–40]. Bases in binding sites that vary from this sequence are indi-
cated in lower case. The sequences are for chicken chkYB-1b/2 pro-
teins [41], for Xenopus FRGY1/2 proteins derived using the selex
procedure [61], for dbpB/YB-1 binding to CD44 pre-mRNA [47],
dbpB/YB-1 binding to the IL-2 5¢-UTR [31] and the VEGF 5¢-UTR
CSD site 1. MSY1, ChkYB-1b and FRGY1 are dbpB/YB-1 proteins.
MSY-4 and chkYB-2 are dbpA proteins, and MSY2 and FRGY2 are
germ cell-specific CSD proteins.
656 L. S. Coles et al. (Eur. J. Biochem. 271) Ó FEBS 2004
sequences has not previously been reported. The ability of
CSD proteins to recognize both types of sequence suggests

that CSD proteins may direct formation of the CSD/PTB
cytoplasmic complexes on the VEGF 5¢-and 3¢-UTR.
Functional role of VEGF CSD/PTB binding sites
The stability of VEGF mRNA is increased by stress
conditions such as hypoxia [7,11–14,17,18] in response to
a number of signaling pathways [12,14,16]. Investigation of
stabilization mechanisms in noninduced or normoxic con-
ditions has not previously been reported. Our data presen-
ted here suggests that the VEGF CSD/PTB sites may be
involved in such mechanisms. We observed, that deletion of
the VEGF 5¢-and3¢-UTR CSD/PTB site sequences, results
in reduced VEGF mRNA stability in both normoxic and
hypoxic conditions while the degree of stabilization of the
VEGF mRNA under hypoxic conditions was not affected.
It appears therefore that CSD/PTB complexes may be
playing a general protective role for the VEGF mRNA in
normoxic growing cells, but that they are not involved in
increased stabilization in response to hypoxia. Consistent
with this finding, CSD proteins have been shown to play a
role in both induced [31–33] and general [34–37] mRNA
stabilization. Recent data suggests that PTB proteins may
also play a role in this latter type of mRNA stabilization
[62]. Factors such as HuR and hnRNPL proteins, have been
implicated in VEGF mRNA stabilization through binding
to the 3¢-UTR, but this is the first report of identification of
potential post-transcriptional regulatory factors binding to
the VEGF 5¢-UTR [17,18]. CSD and PTB proteins may
function to stabilize structures required to enhance mRNA
stability, as proposed for other post-transcriptional roles
mediated by CSD and PTB proteins [42,45]. As the second

5¢-UTR CSD/PTB site is downstream of a reported
alternative transcription start site, the presence of two
CSD/PTB sites in the VEGF 5¢-UTR may be to ensure that
at least one of these sites will be present in alternative forms
of the VEGF mRNA [63].
Given that CSD and CSD-related proteins can play a role
in both cap-dependent [26,27,42] and IRES-driven transla-
tion [43–46] it is possible that the CSD/PTB sites play a role
in translation as well as stabilization of the VEGF mRNA.
A combined role in mRNA stability and translation has
been observed for the YB-1 and MSY-2 CSD proteins
[35–37]. Sequence-specific CSD binding sites involved in
translational regulation have been found in both 5¢-and
3¢-UTR sequences [38–41]. PTB proteins are also involved
in translational regulation and it has been demonstrated
that PTB in combination with a CSD-related protein,
UNR, is involved in IRES function [43–46]. The CSD/PTB
sites are, however, outside the regions defined for IRES
activity in the mouse and human VEGF 5¢-UTR sequences
[19–21], hence a cap-dependent translational role would be
more likely for the VEGF CSD/PTB binding regions.
Higher order CSD/PTB complex formation
We have shown that the VEGF 5¢-UTR CSD/PTB-
containing complexes are multiprotein complexes contain-
ing proteins of the appropriate size for PTB (60 kDa), the
dbpB/YB-1 CSD protein (50 kDa) and two additional
smaller unidentified factors (27 and 12 kDa), giving a
combined molecular mass  180 kDa. DbpB/YB-1 (also
called MSY-1, chkYb-1b, p50 and FRGY1) is one of two
ubiquitously expressed CSD proteins. The other being

represented by dbpA (also called MSY-4, chkYB-2 and
YB2/RYBa) [22–25,29]. Both CSD and PTB proteins have
been shown to functionally interact with a number of
partner proteins e.g. [23,43] but the dbpA or dbpB/YB-1
CSD proteins have not previously been reported to
complex, or interact, with PTB. Although not investigated
here, a role for dbpA in the CSD/PTB complexes can not
be ruled out, as dbpA and dbpB/YB-1 have very similar
functions [38–41]. The large CSD-related UNR protein,
with a size of 97 kDa is, however, unlikely to be part of the
complex.
Our data also demonstrates that the VEGF 5¢-UTR
cytoplasmic CSD/PTB complex is preformed in solution.
This indicates that the preformed CSD/PTB complex, with
an approximate size of 400–490 kDa, may contain addi-
tional proteins to those detected by UV cross-link analysis
(the combined size of cross-linked components is only
180 kDa). Alternatively the 400–490 kDa complex may
contain two molecules of each protein identified by UV
cross-linking. This is likely as CSD and PTB proteins have
in fact been found to be able to bind to RNA and DNA as
dimers as well as monomers [28,29,42,51,56].
Comparison of the sequence requirements for CSD/PTB
complex formation with the binding of recombinant dbpB/
YB-1 and PTB, suggests that CSD proteins may direct the
binding of the multiprotein complex to the VEGF RNA, as
discussed above. Consistent with this, using recombinant
proteins, we have found that dbpB/YB-1 enhances the
binding of PTB to VEGF RNA (M. A. Bartley & L. S. Cole,
unpublished observation). Similarly the CSD-related

protein, UNR, has been shown to direct the binding of
PTB to IRES sequences [45]. CSD proteins have also been
shown to affect the ability of transcription factors to bind to
DNA [23].
Broader role for CSD/PTB complexes in growth factor
gene regulation
We found that the CSD/PTB complexes binding to the
VEGF 5¢-UTR bound to similar sequences in the IL-2
5¢-UTR (Fig. 8B). The IL-2 sequence that binds the CSD/
PTB complex has previously been shown to be part of a
stability element that responds to the JNK signaling pathway
in T cells [31]. We confirmed that the CSD/PTB complexes
could in fact form in T cells. Both the dbpB/YB-1 CSD
protein and another RNA binding protein, nucleolin, were
found to be required for the induced IL-2 mRNA stabiliza-
tion [31]. It therefore appears likely that dbpB/YB-1 may be
able to partner with multiple proteins on the IL-2 5¢-UTR
stability element and that these complexes may respond to
different signaling events or operate under different condi-
tions. The CSD/PTB complexes, for example, could be
involved in general protection of the IL-2 mRNA until
appropriate signaling pathways are activated, whereby
CSD proteins could partner with nucleolin to bring about
enhanced stabilization. A similar mechanism could be
occurring on the VEGF mRNA, where CSD/PTB complexes
protecting VEGF mRNA under normal growing conditions
Ó FEBS 2004 CSD and PTB protein complexes on the VEGF mRNA (Eur. J. Biochem. 271) 657
could be replaced by alternative complexes under induced
conditions other than hypoxia. CSD/PTB complexes could
also play a role in general insulin mRNA stability, as PTB has

recently been implicated in insulin mRNA stabilization in
normoxic conditions [62]. Interestingly, we observed that the
insulin mRNA PTB binding site has an overlapping consen-
sus CSD binding site. PTB, to date, has only been implicated
in stability of insulin and VEGF (this report) mRNAs. A
broad role for CSD/PTB complexes in growth factor gene
regulation, is supported by our finding of these complexes in
a number of different mouse, rat and human cell lines (M. A.
Bartley & L. S. Cole, unpublished observation), and is
consistent with the ubiquitous expression of CSD and PTB
proteins [23,52].
The stability of another growth factor mRNA, that for
GM-CSF, is regulated by CSD proteins [32]. In contrast to
the VEGF and IL-2 genes, CSD proteins were reported
to contact AU-rich sequences in the GM-CSF 3¢-UTR.
Several AU-rich sequences are present in the VEGF
3¢-UTR, the possibility exists that they could also bind
CSD proteins [59]. Different types of CSD binding site may
therefore be targets for alternative types of CSD complexes
responding to different signaling.
Model for combined transcriptional and
post-transcriptional regulation of VEGF expression
We have previously reported that the VEGF and GM-CSF
genes are transcriptionally regulated by CSD proteins
[49–51]. In both cases, CSD proteins act as transcriptional
repressors and our data suggest that the CSD proteins are
removed upon appropriate induction to allow full promoter
activity. Consistent with this, we have found that CSD
proteins bind to a repressor element in the IL-2 gene
(L. S. Cole, unpublished data). CSD proteins may therefore

play a combined transcriptional and post-transcriptional
role in regulation of certain growth factor genes. It seems
likely then that CSD proteins could bind to RNA in the
nucleus at the time of transcription. In support of this, CSD
proteins have in fact been shown to bind to RNA
concomitant with transcription [64]. Furthermore, CSD
proteins have been found to be involved in RNA packaging
and transport to the cytoplasm [26,28,29]. It is feasible that
CSD proteins could be involved at every level of growth
factor gene regulation, from transcription in the nucleus to
transport to the cytoplasm, mRNA stability and translation.
Acknowledgements
We thank Tom Cooper (Houston, Texas) and Doug Black (UCLA) for
gifts of PTB expression plasmids and anti-PTB monoclonal antibodies,
respectively. This work was supported by a Heart Foundation
Australia project grant (G. J. G.), a Cancer Council of South Australia
project grant (L. S. C.), an RAH/IMVS project grant (L. S. C) and a
National Health and Medical Research Committee program grant
(M.A.V,G.J.G.).
References
1. Yancopoulos, G.D., Davis, S., Gale, N.W., Rudge, J.S., Wiegand,
S.J. & Holash, J. (2000) Vascular-specific growth factors and
blood vessel formation. Nature 407, 242–248.
2. Neufeld, G., Cohen, T., Gengrinovitch, S. & Poltorak, Z. (1999)
Vascular endothelial growth factor (VEGF) and its receptors.
FASEB J. 13, 9–22.
3. Veikkola, T. & Alitalo, K. (1999) VEGFs, receptors and angio-
genesis. Semin. Cancer Biol. 9, 211–220.
4. Baumgartner, I. & Isner, J.M. (2001) Somatic gene therapy in the
cardiovascular system. Annu. Rev. Physiol. 63, 427–450.

5. Freedman, S.B. & Isner, J.M. (2001) Therapeutic angiogen-
esis for ischemic cardiovascular disease. J. Mol. Cell. Cardiol. 33,
379–393.
6. Hammond, H.K. & Mckirnan, M.D. (2001) Angiogenic gene
therapy for heart disease: a review of animal studies and clinical
trials. Cardiovasc. Res. 49, 561–567.
7. Goodall, G.J., Coles, L.S., Bartley, M.A. & Lang, K.J.D. (2002)
Post-transcriptional regulation of VEGF. In Genetics of Angio-
genesis (Hoying, J.B., ed.), pp. 69–83. BIOS Scientific Publishers
Ltd, Oxford.
8. Fukumura, D., Xavier, R., Sugiura, T., Chen, Y., Park, E.C., Lu,
N.,Selig,M.,Nielsen,G.,Taksir,T.,Jain,R.K.&Seed,B.(1998)
Tumor induction of VEGF promoter activity in stromal cells. Cell
94, 715–725.
9. Zhao, L. & Eghbali-Webb, M. (2001) Release of pro- and anti-
angiogenic factors by human cardiac fibroblasts. Biochim. Bio-
phys. Acta 1538, 273–282.
10. Balasubramanian, S., Ramakrishnan, S., Charboneau, R., Wang,
J., Barke, R.A. & Roy, S. (2001) Morphine sulfate inhibits
hypoxia-induced vascular endothelial growth factor expression in
endothelial cells and cardiac myocytes. J. Mol. Cell. Cardiol. 33,
2179–2187.
11. Tanaka, T., Kanai, H., Sekiguchi, K., Aihara, Y., Yokoyama, T.,
Arai, M., Kanda, R. & Kurabayashi, M. (2000) Induction of
VEGF gene transcription by IL-1beta is mediated through stress-
activated MAP kinases and SP1 sites in cardiac myocytes. J. Mol.
Cell. Cardiol. 32, 1955–1967.
12. Pages,G.,Berra,E.,Milanini,J.,Levy,A.P.&Pouyssegur,J.
(2000) Stress-activated protein kinases (JNK and p38/HOG) are
essential for vascular endothelial growth factor mRNA stability.

J. Biol. Chem. 275, 26484–26491.
13. Iida, K., Kawakami, Y., Sone, H., Suzuki, H., Yatoh, S., Isobe,
K., Takekoshi, K. & Yamada, N. (2002) Vascular endothelial
growth factor gene expression in a retinal pigmented cell is
up-regulated by glucose deprivation through 3¢UTR. Life Sci. 71,
1607–1614.
14. Shih, S C., Mullen, A., Abrams, K., Mukhopadhyay, D. &
Claffey, K.P. (1999) Role of protein kinase C isoforms in
phorbol ester-induced vascular endothelial growth factor
expression in human glioblastoma cells. J. Biol. Chem. 274,
15407–15414.
15. Dibbens, J.A., Miller, D.L., Damert, A., Risau, W., Vadas, M.A.
& Goodall, G.J. (1999) Hypoxic regulation of vascular endothelial
growth factor mRNA stability requires the cooperation of multi-
ple RNA elements. Mol. Biol. Cell 10, 907–919.
16. Levy,A.P.,Levy,N.S.&Goldberg,M.A.(1996)Post-transcrip-
tional regulation of vascular endothelial growth factor by hypoxia.
J. Biol. Chem. 271, 2746–2753.
17. Shih, S C. & Claffey, K.P. (1999) Regulation of human vascular
endothelial growth factor mRNA stability in hypoxia by hetero-
geneous nuclear ribonucleoprotein L. J. Biol. Chem. 274, 1359–
1365.
18. Goldberg-Cohen, I., Furneaux, H. & Levy, A.P. (2002) A 40-bp
element that mediates stabilization of vascular endothelial growth
factor mRNA by HuR. J. Biol. Chem. 277, 13635–13640.
19. Miller, D.L., Dibbens, J.A., Damert, A., Risau, W., Vadas, M.A.
& Goodall, G.J. (1998) The vascular endothelial growth factor
mRNA contains an internal ribosome entry site. FEBS Letts. 434,
417–420.
658 L. S. Coles et al. (Eur. J. Biochem. 271) Ó FEBS 2004

20. Stein,I.,Itin,A.,Eina,P.,Skaliter,R.,Grossman,Z.&Keshet,E.
(1998) Translation of vascular endothelial growth factor mRNA
by internal ribosme entry: implications for translation under
hypoxia. Mol. Cell. Biol. 18, 3112–3119.
21.Huez,I.,Creancier,L.,Audigier,S.,Gensac,M.,Prats,A.&
Prats, H. (1998) Two independent internal ribosome entry sites are
involved in translation initiation of vascular endothelial growth
factor mRNA. Mol. Cell. Biol. 18, 6178–6190.
22. Shannon, M.F., Coles, L.S., Vadas, M.A. & Cockerill, P.N. (1997)
Signals for activation of the GM-CSF promoter and enhancer in
T cells. Crit. Rev. Immunol. 17, 301–323.
23. Shannon, M.F., Coles, L.S., Attema, J. & Diamond, P. (2001) The
role of architectural transcription factors in cytokine gene tran-
scription. J. Leukoc. Biol. 69, 21–32.
24. Wolffe, A.P. (1994) Structural and functional properties of the
evolutionarily ancient Y-box family of nucleic acid binding pro-
teins. Bioessays 16, 245–251.
25. Graumann, P.L. & Marahiel, M.A. (1998) A superfamily of
proteins that contain the cold-shock domain. Trends Biochem. Sci.
23, 286–290.
26. Matsumoto, K. & Wolffe, A.P. (1998) Gene regulation by Y-box
proteins: coupling control of transcription and translation. Trends
Cell Biol. 8, 318–323.
27. Sommerville, J. (1999) Activities of cold-shock domain proteins in
translation control. Bioessays 21, 319–325.
28. Sommerville, J. & Ladomery, M. (1996) Masking of mRNA by
Y-box proteins. FASEB J. 10, 435–443.
29. Wolffe, A.P., Tafuri, S., Ranjan, M. & Familari, M. (1992) The
Y-box factors: a family of nucleic acid binding proteins conserved
from Escherichia coli to man. New Biol. 4, 290–298.

30. Doniger,J.,Landsman,D.,Gonda,M.A.&Wistow,G.(1992)
The product of unr, the highly conserved gene upstream of
N-ras, contains multiple repeats similar to the cold-shock
domain (CSD), a putative DNA-binding motif. New Biol. 4,
389–395.
31. Chen,C.Y.,Gherzi,R.,Andersen,J.S.,Gaietta,G.,Jurchott,K.,
Royer, H.D., Mann, M. & Karin, M. (2000) Nucleolin and YB-1
are required for JNK-mediated interleukin-2 mRNA stabilization
during T-cell activation. Genes Dev. 14, 1236–1248.
32. Capowski, E., Esnault, S., Bhattacharya, S. & Malter, J. (2001) Y
box-binding factor promotes eosinophil survival by stabilizing
granulocyte-macrophage colony-stimulating factor mRNA.
J. Immunol. 167, 5970–5976.
33. Skalweit, A., Doller, A., Huth, A., Kahne, T., Persson, P.B. &
Thiele, B J. (2003) Posttranscriptional control of renin synthesis:
identification of proteins interacting with renin mRNA-3¢-
untranslated region. Circ. Res. 92, 419–427.
34. Iuchi, Y., Kaneko, T., Matsuki, S., Sasagawa, I. & Fujii, J. (2003)
Concerted changes in the YB-2/RYB-a protein and protamine 2
messenger RNA in the mouse testis under heat stress. Biol. Reprod.
68, 129–135.
35. Yu, J., Hecht, N.B. & Schultz, R.M. (2001) Expression of MSY2
in mouse oocytes and preimplantation embryos. Biol. Reprod. 65,
1260–1270.
36. Yu, J., Hecht, N.B. & Schultz, R.M. (2002) RNA-binding prop-
erties and translation repression in vitro by germ cell-specific
MSY2 protein. Biol. Reprod. 67, 1093–1098.
37. Evdokimova, V., Ruzanov, P., Imataka, H., Raught, B., Svitkin,
Y., Ovchinnikov, L.P. & Sonenberg, N. (2001) The major mRNA-
associated protein YB-1 is a potent 5¢-cap-dependent mRNA

stabilizer. EMBO J. 20, 5491–5502.
38. Braun, R.E. (2000) Temporal control of protein synthesis during
spermatogenesis. Int. J. Androl. 23, 92–94.
39. Davies, H.G., Giorgini, F., Fajardo, M.A. & Braun, R.E. (2000) A
sequence-specific RNA binding complex expressed in murine germ
cells contains MSY2 and MSY4. Dev. Biol. 221, 87–100.
40. Giorgini, F., Davies, H.G. & Braun, R.E. (2001) MSY2 and MSY4
bind a conserved sequence in the 3¢ untranslated region of pro-
tamine 1 mRNA in vitro and in vivo. Mol. Cell Biol. 21, 7010–7019.
41. Swamynathan, S.K., Nambiar, A. & Guntaka, R.V. (2000)
Chicken Y-box proteins chk-YB-1b and chk-YB-2 repress trans-
lation by sequence–specific interaction with single-stranded RNA.
Biochem. J. 348, 297–305.
42. Evdokimova, V.M. & Ovchinnikov, L.P. (1999) Translational
regulation by Y-box transcription factor: involvement of the
major mRNA-associated protein, p50. Int. J. Biochem. Cell Biol.
31, 139–149.
43. Hunt, S.L., Skern, T., Liebig, H.D., Kuechler, E. & Jackson, R.J.
(1999) Rhinovirus 2A proteinase mediated stimulation of rhino-
virus RNA translation is additive to the stimulation effected by
cellular RNA binding proteins. Virus Res. 62, 119–128.
44. Hunt, S.L., Hsuan, J.J., Totty, N. & Jackson, R.J. (1999) unr, a
cellular cytoplasmic RNA-binding protein with five cold-shock
domains, is required for internal initiation of translation of human
rhinovirus RNA. Genes Dev. 13, 437–448.
45. Mitchell, S.A., Brown, E.C., Coldwell, M.J., Jackson, R.J. & Willis,
A.E. (2001) Protein factor requirements of the Apaf-1 internal
ribosome entry segment: roles of polypyrimidine tract binding
protein and upstream of N-ras. Mol. Cell. Biol. 21, 3364–3374.
46. Mitchell, S.A., Spriggs, K.A., Coldwell, M.J., Jackson, R.J. &

Willis, A.E. (2003) The Apaf-1 internal ribosome entry segment
attains the correct structural conformation for function via inter-
actions with PTB and unr. Mol. Cell 11, 757–771.
47. Stickeler, E., Fraser, S.D., Honig, A., Chen, A.L., Berget, S.M. &
Cooper, T.A. (2001) The RNA binding protein YB-1 binds
A/C-rich exon enhancers and stimulates splicing of the CD44
alternative exon v4. EMBO J. 20, 3821–3830.
48. Chansky, H.A., Hu, M., Hickstein, D.D. & Yang, L. (2001)
Oncogenic TLS/ERG and EWS/Fli-1 fusion proteins inhibit
RNA splicing mediated by YB-1 protein. Cancer Res. 61, 3586–
3590.
49. Coles, L.S., Diamond, P., Occhiodoro, F., Vadas, M.A. &
Shannon, M.F. (2000) An ordered array of cold shock domain
repressor elements across tumor necrosis factor-responsive ele-
ments of the granulocyte-macrophage colony-stimulating factor
promoter. J. Biol. Chem. 275, 14482–14493.
50. Coles, L.S., Diamond, P., Occhiodoro, F., Vadas, M.A. &
Shannon, M.F. (1996) Cold shock domain proteins repress tran-
scription from the GM-CSF promoter. Nucleic Acids Res. 24,
2311–2317.
51. Coles,L.S.,Diamond,P.,Lambrusco,L.,Hunter,J.,Burrows,J.,
Vadas, M.A. & Goodall, G.J. (2002) A novel mechanism of
repression of the vascular endothelial growth factor promoter, by
single strand DNA binding cold shock domain (Y-box) proteins in
normoxic fibroblasts. Nucleic Acids Res. 30, 4845–4854.
52. Brunel, F., Alzari, P.M., Ferrara, P. & Zakin, M.M. (1991)
Cloning and sequencing of PYBP, a pyrimidine-rich specific single
strand DNA-binding protein. Nucleic Acids Res. 19, 5237–5245.
53. Brunel, F., Zakin, M.M., Buc, H. & Buckle, M. (1996) The
polypyrimidine tract binding (PTB) protein interacts with single-

stranded DNA in a sequence-specific manner. Nucleic Acids Res.
24, 1608–1615.
54. Yuan, X., Davydova, N., Conte, M.R., Curry, S. & Matthews, S.
(2002) Chemical shift mapping of RNA interactions with the
polypyrimidine tract binding protein. Nucleic Acids Res. 30,
456–462.
55. Conte, M.R., Grune, T., Ghuman, J., Kelly, G., Ladas, A.,
Matthews, S. & Curry, S. (2000) Structure of tandem RNA
recognition motifs from polypyrimidine tract binding protein
reveals novel features of the RRM fold. EMBO J. 19, 3132–3141.
56. Oh, Y.L., Hahm, B., Kim, Y.K., Lee, H.K., Song, O.K., Tsu-
kiyama-Kohara, K., Kohara, M., Nomoto, A. & Jang, S.K.
Ó FEBS 2004 CSD and PTB protein complexes on the VEGF mRNA (Eur. J. Biochem. 271) 659
(1998) Determination of functional domains in polypyrimidine-
tract-binding protein. Biochem. J. 331, 169–175.
57. Perez, I., Lin, C.H., McAfee, J.G. & Patton, J.G. (1997) Mutation
of PTB binding sites causes misregulation of alternative 3¢ splice
site selection in vivo. RNA 3, 764–778.
58. Shima, D.T., Kuroki, M., Deutsch, U., Ng, Y.S., Adamis, A.P. &
D’Amore, P.A. (1996) The mouse gene for vascular endothelial
growth factor: genomic structure, definition of the transcriptional
unit, and characterization of transcriptional and post-transcrip-
tional regulatory sequences. J. Biol. Chem. 271, 3877–3883.
59. Dibbens,J.A.,Polyak,S.W.,Damart,A.,Risau,W.,Vadas,M.A.
& Goodall, G.J. (2001) Nucleotide sequence of the mouse VEGF
3¢UTR and quantitative analysis of sites of polyadenylation.
Biochim. Biophys. Acta 93512, 1–6.
60. Schreiber, E., Matthias, P., Muller, M.M. & Schaffner, W.
(1989) Rapid detection of octamer binding proteins with Ômini-
extractsÕ, prepared from a small number of cells. Nucleic Acids Res.

17, 6419.
61. Bouvet, P., Matsumoto, K. & Wolffe, A.P. (1995) Sequence-
specific RNA recognition by the Xenopus Y-box proteins: an
essential role for the cold shock domain. J. Biol. Chem. 270,
28297–28303.
62. Tillmar, L. & Welsh, N. (2002) Hypoxia may increase rat insulin
mRNA levels by promoting binding of the polypyrimidine tract-
binding protein (PTB) to the pyrimidine-rich insulin mRNA-3¢
untranslated region. Mol. Med. 8, 263–272.
63. Akiri, G., Nahari, D., Finkelstein, Y., Le, S Y., Elroy-Stein, O. &
Levi, B Z. (1998) Regulation of vascular endothelial growth fac-
tor (VEGF) expression is mediated by internal initiation of
translation and alternative initiation of transcription. Oncogene
17, 227–236.
64. Soop, T., Nashchekin, D., Zhao, J., Sun, X., Alzhanova-Ericsson,
A.T., Bjorkroth, B., Ovchinnikov, L. & Daneholt, B. (2003) A
p50-like Y-box protein with a putative translational role becomes
associated with pre-mRNA concomitant with transcription.
J. Cell Sci. 116, 1493–1503.
660 L. S. Coles et al. (Eur. J. Biochem. 271) Ó FEBS 2004