Báo cáo khoa học: Regulation of ascidian Rel by its alternative splice variant pdf

Regulation of ascidian Rel by its alternative splice variant
Narudo Kawai
1
, Masumi Shimada
1
, Hiroyuki Kawahara
1
, Noriyuki Satoh
2
and Hideyoshi Yokosawa
1
1
Department of Biochemistry, Graduate School of Pharmaceutical Sciences, Hokkaido University, Sapporo 060-0812, Japan;
2
Department of Zoology, Graduate School of Science, Kyoto University, Kyoto 606-8502, Japan
The Rel/NF-jB family of transcription factors play key roles
in morphogenesis and immune responses. We reported
previously that As-rel1 and As-rel2 of the ascidian Halo-
cynthia roretzi are involved in notochord formation. The
As-rel1 protein is a typical Rel/NF-jB family member,
whereas the As-rel2 protein is a novel truncated product
of As-rel1 that lacks a nuclear localization signal and the
unique C-terminal region. Here, we present conclusive evi-
dence that As-rel1 and As-rel2 are generated from a single
gene by alternative splicing. We analyzed the roles of As-rel2
using cells transfected with As-rel1 or As-rel2 or both.
As-rel1 was localized in the nucleus and As-rel2 in the
cytoplasm when they were transfected individually. In con-
trast, when they were transfected simultaneously, both were
localized in the nucleus because of the association of As-rel2
with As-rel1. In this case, the transcriptional activity of

As-rel1 was suppressed by As-rel2. Ascidian IjB was found
to sequester As-rel1 in the cytoplasm and suppress its tran-
scriptional activity when As-rel1 and IjB were transfected
simultaneously. In contrast, when As-rel1 and IjBwere
transfected together with As-rel2, As-rel1 was transported
into the nucleus and its transcriptional activity was rescued
from inhibition by IjB, whereas As-rel2 remained localized
in the cytoplasm, suggesting IjB sequestration in the cyto-
plasm by As-rel2. From these findings, we conclude that the
alternative splice variant, As-rel2, regulates the nuclear
localization and transcriptional activity of As-rel1.
Keywords: alternative splicing; IjB; nuclear transport; Rel;
transcriptional activity.
The Rel/NF-jB family of transcription factors contains the
highly conserved Rel homology domains (RHDs) required
for DNA binding and dimerization, and these transcription
factors regulate the expression of downstream target genes
involved in various phenomena such as morphogenesis,
immune response, cell growth, and programmed cell death
[1–8]. Regulatory mechanisms underlying the functions of
Rel/NF-jB family members have been extensively studied,
and several modes of their regulation have been proposed
[5–8]. A suggested basic regulatory mechanism in the
canonical Rel/NF-jB signaling pathway is based on the
inhibitor protein IjB-mediated mechanism. Without extra-
cellular stimuli, the Rel/NF-jB protein is sequestered in the
cytoplasm through binding to IjB because IjBmasksthe
nuclear localization signal of Rel/NF-jB. In response to
extracellular stimuli, the IjB protein is phosphorylated
by the oligomeric kinase IKKb, polyubiquitinated, and

degraded by the proteasome, resulting in the release of the
Rel/NF-jB protein, which moves from the cytoplasm to the
nucleus, where it binds to regulatory elements of target genes
to trigger their transcription. In contrast, in the noncanonical
pathway, based on proteolytic processing of a mammalian
Rel/NF-jB precursor, the precursor p100 binds with RelB
to form a heterodimer in the cytoplasm. In response to
extracellular stimuli, the p100 protein in the complex is
phosphorylated by the oligomeric kinase IKKa and conver-
ted into p52 via the ubiquitin-mediated proteolytic pathway.
The RelB/p52 heterodimer thus formed then moves to the
nucleus and functions in transcription of target genes. In
addition, post-translational modification of Rel/NF-jB
family members such as phosphorylation [9–11] or small
ubiquitin-related modifier-1 modification [12] is an addi-
tional regulatory mechanism for their transcriptional acti-
vity, the phosphorylated protein thus showing enhanced
transcriptional activity. On the other hand, the function of
the Rel/NF-jB protein is negatively regulated by its
C-terminally truncated form. For example, a homodimer
of p50, a mammalian Rel/NF-jB family member that lacks a
C-terminal transactivation domain, binds to the jB consen-
sus element and exhibits a suppressive effect on jB-depend-
ent gene expression [13]. A C-terminally truncated form of
p65, which is produced by caspase-catalyzed cleavage upon
apoptosis, has the nuclear localization signal but no
transcriptional activity. This protein plays a suppressive role
in wild-type p65-mediated transcription [14].
In a previous study [15], we found that the ascidian
Halocynthia roretzi has two members of the Rel/NF-jB

family: one called As-rel1 is a typical Rel/NF-jB family
member that has RHD, the nuclear localization sequence,
and the C-terminal region; the other called As-rel2 has a
novel structure with deletion of both the nuclear localization
signal and the C-terminal region. In H. roretzi embryos, it
was found that the ectopic expression of As-rel1 protein led
to reduction of the number of notochord cells and a defect
Correspondence to H.Yokosawa,DepartmentofBiochemistry,
Graduate School of Pharmaceutical Sciences, Hokkaido University,
Sapporo 060-0812, Japan.
Fax: + 81 11 706 4900, Tel.: + 81 11 706 3754,
E-mail:
Abbreviations: RHD, Rel homology domain; IKK, IjBkinase;
GST, glutathione S-transferase; EST, expressed sequence tag;
DAPI, 4¢,6-diamidino-2-phenylindol; GFP, green fluorescent protein.
(Received 20 June 2003, revised 24 August 2003,
accepted 18 September 2003)
Eur. J. Biochem. 270, 4459–4468 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03838.x
in tail elongation, indicating that this Rel/NF-jBprotein
affects notochord formation in ascidian embryos. On the
other hand, it was found that the As-rel2 protein had an
antagonistic effect on the action of the As-rel1 protein. These
findings imply that As-rel2, the C-terminally truncated form
of As-rel1, regulates the function of As-rel1. However, the
regulatory mechanism has not yet been elucidated.
To determine the relationship between As-rel1 and
As-rel2, we investigated the effects of As-rel2 on the
localization and activity of As-rel1 by using As-rel1-
transfected and As-rel2-transfected mammalian cells.
As-rel2, transfected alone, was localized in the cytoplasm

because of the lack of the nuclear localization sequence,
whereas this protein when transfected together with As-rel1
moved to the nucleus and suppressed transcriptional activity
of As-rel1, possibly because the level of activity of a
heterodimer of As-rel1 and As-rel2 is lower than that of a
homodimer of As-rel1. On the other hand, As-rel1 trans-
fected together with the ascidian inhibitor IjBremained
localized in the cytoplasm, but this protein transfected
together with both IjB and As-rel2 was able to move into
the nucleus, and its transcriptional activity was rescued from
inhibition by IjB. These results indicate that As-rel2
regulates the function of As-rel1 in the presence of IjB.
We presented conclusive evidence that As-rel1 and As-rel2
are generated from a single gene by alternative splicing. To
the best of our knowledge, this is the first report of a novel
regulatory mechanism for the function of the Rel/NF-jB
family that is mediated by an alternative splice variant.
Materials and methods
Materials
The proteasome inhibitor MG132 was purchased from the
Peptide Institute, Inc. (Osaka, Japan). Mouse anti-(T7-tag),
anti-(Flag-tag) M2 and anti-(c-Myc-tag) 9E10 monoclonal
Igs were purchased from Novagen, Sigma and Santa Cruz,
respectively. Horseradish peroxidase-conjugated antibodies
against rabbit and mouse immunoglobulins were obtained
from Amersham Pharmacia Biotech.
Preparation of antibody against As-rel1
A 480-bp fragment, consisting of 460–618 amino-acid
residues of As-rel1, was subcloned in-frame into the
pGEX-6P-1 expression vector with the correct orientation.

Escherichia coli BL21 expressing glutathione S-transferase
(GST)–As-rel1 fusion protein after isopropyl b-
D
-thio-
galactoside induction was solubilized in lysis buffer
(20 m
M
Tris/HCl, pH 8.0, containing 1 m
M
EDTA and
100 m
M
NaCl), and the extract was applied to glutathione-
immobilized agarose beads to trap the fusion protein, which
was subsequently digested with PreScission protease (Amer-
sham) to release the As-rel1 fragment. The isolated As-rel1
fragment was emulsified in Freund’s complete adjuvant and
injected subcutaneously into rabbits followed by a four-time
booster injection of the isolated fragment in Freund’s
incomplete adjuvant. The antiserum was purified by affinity
chromatography on Protein A-immobilized Sepharose
CL-4B (Amersham), and the immunoreactivity was verified
by Western blotting of the fusion protein.
Preparation of the genome and genomic PCR
The genome of the ascidian, H. roretzi, was prepared from
muscles. A 5 mL volume of extraction buffer [2 · lysis
buffer (ABI), 0.5 · phosphate-buffered saline, and
200 lgÆmL
)1
proteinase K (ABI)] was added to 0.5 g

muscle, and then incubated overnight at 55 °C. Phenol
(5 mL) was added to the suspension, and the mixture was
rotated for 1 h at room temperature and was then
centrifuged at 10 000 g for 10 min. The resulting super-
natant was mixed with phenol, rotated for 8 h at room
temperature and centrifuged at 10 000 g for 10 min. The
resulting supernatant was subjected to ethanol precipitation
to give genomic DNA as a precipitate.
Genomic PCR was performed with several combinations
of the following primers for the As-rel sequence [genome
forward F1 primer, 5¢-ATGGACAAAATGTCTA-3¢
(As-rel1 bp 47–62); genome forward F2 primer, 5¢-GGA
AGCCACAAAAGTTAT-3¢ (As-rel1 bp 832–849); gen-
ome forward F3 primer, 5¢-CTGCTGGATAGATGA
TGC-3¢ (As-rel2 bp 941–958); genome forward F4 primer,
5¢-ACTTATTTTTCTCGCACA-3¢ (As-rel2 bp 1700–1717);
genome reverse R1 primer, 5¢-TCTGATTGAGGTTAG
TGG-3¢ (As-rel2 bp 1580–1563); genome reverse R2
primer, 5¢-TTTCGGTTTGTAATGTTAGT-3¢ (As-rel2
bp 1656–1647); genome reverse R3 primer, 5¢-GAATA
CGAACCCAAACAA-3¢ (As-rel2 bp 2117–2100); genome
reverse R4 primer, 5¢-TGTCTTCATGTGGGACAA-3¢
(As-rel1 bp 1441–1434)] using an Expand High Fidelity
Taq polymerase (Roche). PCR fragments produced were
cloned into the pGEM-T-vector (Promega), and the
sequences were determined from each end of insert DNAs.
Cell culture and transfection
HEK293 or 293T cells were cultured in Dulbecco’s modified
Eagle’s medium containing 10% fetal bovine serum at
37 °C under a 5% CO

2
atmosphere. Transfection was
performed using Effectene transfection reagent (Qiagen),
Metafectene transfection reagent (Biontex), or FuGene 6
transfection reagent (Roche) according to the manufac-
turer’s protocol.
Isolation of an IjB homologue
We searched the Ciona intestinalis expressed sequence tag
(EST) database (Ciona cDNA resources of Kyoto Univer-
sity) for an IjB homologue and found it in this database.
The EST clone, cieg29i06, which is predicted to include the
full length of IjB-like cDNA, was subjected to nucleotide
sequence determination, and this clone, designated Ci-IjB
for an IjB homologue of the ascidian C. intestinalis,was
found to include the full length.
Plasmid constructions
To generate As-rel1 and As-rel2 expression plasmids,
pCI-neo-As-rel1, pCI-neo-Flag-As-rel1, pCI-neo-c-Myc-
As-rel1, pCI-neo-c-6Myc-As-rel1, pCI-neo-Flag-As-rel2,
pCI-neo-c-6Myc-As-rel2 and pCI-neo-T7-As-rel2, full-
length As-rel1 and As-rel2 cDNAs were inserted into the
MluI–NotI sites of pCI-neo (Promega), pCI-neo-Flag,
4460 N. Kawai et al.(Eur. J. Biochem. 270) Ó FEBS 2003
pCI-neo-c-Myc, pCI-neo-c-6Myc and pCI-neo-T7, respect-
ively, after subcloning of full-length As-rel1 and As-rel2
containing the MluIandNotI sites by PCR using the
following primers (As-rel MulI forward primer, 5¢-ACGC
GTATGGACAAAATGTCTA-3¢; As-rel1 NotI reverse
primer, 5¢-AGCGGCCGCTCAGTTGTAATTC-3¢;As-
rel2 NotI reverse primer, 5¢-AGCGGCCGCTCATCT

ATCCAGCA-3¢). The PCR products were subcloned in
the pGEM-T-vector (Promega). pEGFP-C1-As-rel1 was
constructed by PCR using the following primers (As-rel1
XhoI forward primer, 5¢-CTCGAGCTATTATGGA
CAAAATGT-3¢; As-rel1 BamHI reverse primer, 5¢-GGAT
CCGTTGTAATTCTGAT-3¢). The PCR product sub-
cloned in the pGEM-T-vector was digested with XhoIand
BamHI and then inserted into the XhoIandBamHIsitesof
pEGFP (Clontech). pBIND-As-rel1 and pBIND-As-rel2
were constructed by PCR using the following primers
(As-rel MluI forward primer, 5¢-ACGCGTTGATGGAC
AAAAT-3¢; As-rel1 NotI reverse primer, 5¢-AGCGGCCG
CTCAGTTGTAATTC-3¢;As-rel2NotI reverse primer,
5¢-AGCGGCCGCTCATCTATCCAGCA-3¢). The PCR
products subcloned in the pGEM-T-vector were digested
with MluIandNotI and then inserted into the MluIand
NotI sites of pBIND (Promega). pCI-neo-Flag-Ci-IjBand
pCI-neo-c-6Myc-Ci-IjB were constructed by PCR using the
following primers (Ci-IjB SalI forward primer, 5¢-GTCGA
CATGTCTAATAAAGCA-3¢;Ci-IjB NotI reverse primer
5¢-GCGGCCGCTCATTGTCG-3¢). The PCR products
subcloned in the pGEM-T-vector were digested with SalI
and NotI and then inserted into the SalIandNotIsitesof
pCI-neo-Flag or pCI-neo-c-6Myc.
Immunoblotting
Proteins were separated by SDS/PAGE on a 10% gel and
transferred to a nitrocellulose membrane (Advantec, Tokyo,
Japan). The membrane was blocked with 5% nonfat milk in
phosphate-buffered saline containing 0.1% Tween 20 for
1 h at room temperature, incubated with the primary

antibody at room temperature for 1 h and then with a
horseradish peroxidase-conjugated antibody against rabbit
or mouse immunoglobulin at room temperature for 30 min,
and developed with an enhanced chemiluminescence detec-
tion system (Amersham).
Immunoprecipitation
HEK293 cells were transfected with several combinations of
1.0 lg pCI-neo-c-Myc-As-rel1 and 1.0 lg pCI-neo-Flag-
As-rel1 using Effectene transfection reagent in 100-mm
dishes. (The total amount of plasmid DNA was adjusted to
2.0 lg with an empty vector, pCI-neo.) After 48 h of
incubation, the cells were washed with phosphate-buffered
saline and disrupted by treatment with lysis buffer (50 m
M
Tris/HCl, pH 8.0, containing 150 m
M
NaCl, 0.1% Nonidet
P40, and 10% glycerol) containing 10 l
M
MG132 and a
protease inhibitor cocktail (Roche) for 30 min on ice, and
the lysate was centrifuged at 13 000 g for 20 min. The
resulting supernatant was pretreated with 10 lLPro-
tein G-immobilized agarose (Santa Cruz) at 4 °Cfor
30 min and was then incubated with 2 lg anti-(Flag-tag)
M2 Ig and 10 lL Protein G-immobilized agarose at 4 °C
for 1 h. The beads were washed five times with lysis buffer,
boiled for 5 min in SDS sample buffer, and subjected to
SDS/PAGE and then to Western blotting with anti-(Flag-
tag) M2 or anti-(c-Myc-tag) 9E10 Ig. Alternatively,

HEK293 cells were transfected with several combinations
of 0.2 lg pCI-neo-As-rel1 and 1.8 lgpCI-neo-T7-As-rel2.
Immunoprecipitation with both 2 lg anti-(As-rel1) IgG and
10 lL Protein A-immobilized Sepharose was carried out,
followed by Western blotting with antibody to As-rel1 or
T7-tag.
To determine the relationship between As-rel1, As-rel2
and Ci-IjB, HEK293T cells were transfected with several
combinations of 1.0 lg each of pCI-neo-c-6Myc-As-rel1,
pCI-neo-c-6Myc-As-rel2, and pCI-neo-Flag-Ci-IjBusing
Metafectene transfection reagent in 60-mm dishes. (The
total amount of plasmid DNA was adjusted to 3.0 lg
with empty vectors, pCI-neo-Flag and pCI-neo-c-6Myc.)
Immunoprecipitation with anti-(Flag-tag) M2-immobilized
beads (Sigma) was carried out, and the beads were
washed five times with lysis buffer and eluted with
100 lgÆmL
)1
3 · Flag-peptide (Sigma). The eluate was
subjected to a second immunoprecipitation with both 2 lg
anti-(As-rel1) IgG and 10 lL Protein A-immobilized
Sepharose at 4 °C for 1 h. The beads were washed five
times with lysis buffer and eluted with 0.1
M
glycine
(pH 3.0). The eluates obtained from the first and second
immunoprecipitations were subjected to SDS/PAGE and
then to Western blotting with anti-(Flag-tag) M2 or anti-
(c-Myc-tag) Ig. In addition, to confirm the interaction
between As-rel2 and Ci-IjB, HEK293T cells were

transfected with several combinations of 1.0 lgeachof
pCI-neo-Flag-As-rel2 and pCI-neo-c-6Myc-Ci-IjB. Immu-
noprecipitation with anti-(Flag-tag) M2-immobilized
beads and elution with 3 · Flag-peptide were carried
out as described above, followed by Western blotting with
anti-(Flag-tag) M2 or anti-(c-Myc-tag) Ig.
Localization of As-rel1 and As-rel2
HEK293 cells were transfected with 0.3 lg pEGFP-C1-
As-rel1 and 1.2 lg pCI-neo (empty vector) using Metafec-
tene transfection reagent in 35-mm dishes. After 48 h of
incubation, the cells were washed with phosphate-buffered
saline and fixed with 3.7% parafolmaldehyde in phosphate-
buffered saline for 30 min at room temperature. After being
washed twice with phosphate-buffered saline, the specimen
was incubated with 0.15% Triton X-100 in phosphate-
buffered saline for 5 min at room temperature and washed
again twice with phosphate-buffered saline. The specimen
was then stained with 0.1 mgÆmL
)1
4¢,6-diamidino-2-phenyl-
indol (DAPI) in phosphate-buffered saline for 3 min at
room temperature and washed twice with phosphate-
buffered saline. The fixed cells were examined using a Zeiss
Axiophot 2 microscope equipped with a fluorescein iso-
thiocyanate filter set (488-nm excitation) for green fluores-
cent protein (GFP) visualization and with a UV filter set
(372-nm excitation) for DAPI visualization. For analysis of
the localization of As-rel2, HEK293 cells were transfected
with 0.3 lg pCI-neo-Flag-As-rel2 together with 1.2 lgpCI-
neo (empty vector) using Metafectene transfection reagent.

The cells were incubated with anti-(Flag-tag) M2 Ig as a
primary antibody and subsequently with Alexa FluorÒ 594
Ó FEBS 2003 Regulation of Rel by alternative splice variant (Eur. J. Biochem. 270) 4461
goat anti-mouse IgG (Molecular Probes) as a secondary
antibody and then visualized with a rhodamine isothio-
cyanate filter set (590 nm for excitation) for Alexa. For
analysis of the colocalization of As-rel1 and As-rel2,
HEK293 cells were transfected with 0.3 lgeachpEGFP-
C1-As-rel1 and pCI-neo-Flag-As-rel2 together with 0.9 lg
pCI-neo using Metafectene transfection reagent. The cells
were stained and visualized as described above. For analysis
of the effect of Ci-IjB, the cells were transfected with 0.3 lg
each pEGFP-C1-As-rel1 and pCI-neo-c-6Myc-Ci-IjB
together with or without 0.3 lg pCI-neo-Flag-As-rel2 using
Metafectene transfection reagent, and visualization was
performed as described above.
Luciferase assay
To measure transcriptional activities of As-rel1 and As-
rel2, we used GAL4 fusion proteins, GAL4–As-rel1 and
GAL4–As-rel2, and a plasmid containing GAL4-binding
DNA sequences and a luciferase reporter gene [16].
HEK293 cells were transfected with several combinations
of 0.75 lg of the pG5/luc vector (Promega) and pBIND-
As-rel1, pBIND-As-rel2, or pBIND-As-rel1 plus pCI-
neo-Flag-As-rel2 using Metafectene transfection reagent
in 35-mm dishes. (The total amount of plasmid DNA
was adjusted to 1.5 lgwithanemptyvector,pBINDor
pCI-neo-Flag.) Alternatively, the cells were transfected
with the combination of 0.25 lg of the pG5/luc vector,
pBIND-As-rel1 and pCI-neo-Flag-Ci-IjB with or without

pCI-neo-Flag-As-rel2 using FuGene 6 transfection rea-
gent. (The total amount of plasmid DNA was adjusted
to 0.5 lg with an empty vector, pCI-neo-Flag.) After
48 h of incubation, the transfected cells were washed
with phosphate-buffered saline and disrupted with
Passive Lysis Buffer (Dual-LuciferaseÒ Reporter Assay
System; Promega). Luciferase activity of the resulting
lysate was measured by using a Dual-Luciferase Reporter
Assay System and an AB-2000 luminescencer-PNS (Atto,
Tokyo, Japan). The same experiments were repeated
three times. In addition, to assess the expression levels of
GAL4–As-rel1, Flag–As-rel2, and Flag–Ci-IjB, some of
the transfected cells were subjected to SDS/PAGE and
then to Western blotting with anti-(As-rel1) and anti-
(Flag-tag) M2 Ig.
Results
As-rel2 is a splice variant of As-rel1
The nucleotide sequence encoding RHD of As-rel1 is
completely identical with that of As-rel2 [15], indicating the
possibility that As-rel1 and As-rel2 are splice variants. To
understand how As-rel1 and As-rel2 mRNAs are produced,
we determined the partial genomic sequence (4193 nucleo-
tides) of As-rel using genomic PCR products (Fig. 1).
Comparison of the genomic and cDNA sequences revealed
that the genome has one sequence encoding RHD (876
nucleotides), that the As-rel1-specific sequence is located
2369 bp downstream from the RHD sequence, and that the
intron (2369 nucleotides) for As-rel1 contains a stop codon
and a polyadenylation signal sequence, which are located
24 and 2077 bp downstream, respectively, from the RHD

sequence (Fig. 1A). These results indicate that the As-rel1
mRNA is generated from pre-mRNA by splicing at the
splice sites shown in Fig. 1B, while a short mRNA of
As-rel2 is generated because the intron for As-rel1, which
contains a stop codon 24 bp downstream from its 5¢ end, is
not excised (Fig. 1A). These results indicate that As-rel1 and
As-rel2 are splice variants. With regard to the splicing for
the typical Rel/NF-jB family member As-rel1, it should be
noted that the sequence at the 5¢ end of the intron for
As-rel1 is GC (Fig. 1B). The common sequence for the
corresponding 5¢-splice site is GT, but the sequence GC has
been reported in several species [17,18].
Interaction of As-rel1 with As-rel2
As-rel1 and As-rel2 have identical RHDs, which are
necessary for DNA binding, interaction with IjB, and
dimerization. First, to determine whether As-rel1 interacts
with itself, we transiently coexpressed Flag-tagged and
c-Myc-tagged As-rel1s in HEK293 cells, and the extracts of
transfected cells were subjected to immunoprecipitation
using anti-(Flag-tag) Ig and then to Western blotting with
anti-(c-Myc-tag) Ig to detect interaction (Fig. 2A). As
expected, As-rel1 was found to interact with itself to form
a homodimer. Next, we carried out an experiment to
determine whether As-rel1 interacts with As-rel2. As-rel1-
transfected and T7-tagged As-rel2-transfected cells were
subjected to immunoprecipitation using antibody against
As-rel1 and then to Western blotting with antibodies against
As-rel1 and T7-tag (Fig. 2B). As-rel1 was again found to
interact with As-rel2 to form a heterodimer.
Localization of As-rel1 and As-rel2

As As-rel1 has the nuclear localization signal and As-rel2
does not, it is reasonable to assume that the former can
Fig. 1. As-rel1 and As-rel2 are splice variants. (A) Schematic repre-
sentation of the structures of the As-rel genome and mRNAs of
As-rel1 and As-rel2. (B) Genome sequence of splicing sites. Note that
As-rel1 mRNA is generated by splicing at the sites shown by open
arrowheads, whereas As-rel2 mRNA is generated without excision at
the above sites. The stop codon in the alternative exon for As-rel2
is indicated by a closed arrowhead, and the polyadenylation signal
sequence is indicated by an arrow.
4462 N. Kawai et al.(Eur. J. Biochem. 270) Ó FEBS 2003
move into the nucleus, whereas the latter remains localized
in the cytoplasm. It would be interesting to determine
whether coexpression of As-rel2 modulates the localization
of As-rel1 as As-rel2 rescues the effect of As-rel1 on
notochord formation [15]. To determine their localization,
we first transiently expressed GFP–As-rel1 fusion protein
and Flag-tagged As-rel2 in HEK293 cells individually
(Fig. 3A). As expected, As-rel1 was present in the nucleus,
whereas As-rel2 was in the cytoplasm. On the other hand,
when GFP–As-rel1 and Flag-tagged As-rel2 were coex-
pressed in HEK293 cells, both As-rel1 and As-rel2 were
present in the nucleus (Fig. 3B), strongly suggesting that
As-rel2, possibly as a heterodimer with As-rel1, can move to
the nucleus.
Suppressive effect of As-rel2 on transcriptional
activity of As-rel1
It was found that the heterodimer composed of As-rel1 and
As-rel2 is localized in the nucleus. We next carried out an
experiment to determine whether As-rel2 modulates the

transcriptional activity of As-rel1. As DNA sequences that
bind As-rel1 and As-rel2 have not been determined, we
employed a luciferase assay in HEK293T cells using GAL4–
As-rel fusion proteins and the GAL4-binding DNA
sequence. In the cells transiently expressing GAL4–As-rel1
alone, luciferase activity was enhanced, depending on the
dose of GAL4–As-rel1 protein (Fig. 4A), whereas the
activity was undetectable in the cells transiently expressing
GAL4–As-rel2 alone (Fig. 4B). These results suggest that the
C-terminal domain of As-rel1 is indispensable for transcrip-
tional activity. Interestingly, in the cells transiently expressing
GAL4–As-rel1 and Flag-tagged As-rel2 simultaneously, the
activity of As-rel1 was moderately suppressed by As-rel2
(Fig. 4C, upper panel, luciferase assay). It should be noted
that the expression level of GAL4–As-rel1 remained almost
Fig. 2. Interaction of As-rel1 with As-rel2. (A) HEK293 cells were
transiently transfected with the indicated combinations of Flag-As-rel1
and c-Myc-As-rel1 expression plasmids, and 48 h after transfection the
cell lysates were subjected to immunoprecipitation using anti-(Flag-
tag) Ig and then to Western blotting with anti-(Flag-tag) and anti-
(c-Myc-tag) Igs. (B) HEK293 cells were transiently transfected with
indicated combinations of As-rel1 and T7-As-rel2 expression plasmids,
and 48 h after transfection the cell lysates were subjected to
immunoprecipitation (IP) using an antibody against As-rel1 and then
to Western blotting with antibodies against As-rel1 and T7-tag.
Fig. 3. Localization of As-rel1 and As-rel2. (A) HEK293 cells were individually transiently transfected with GFP–As-rel1 (a, b) and Flag–As-rel2
expression plasmids (c, d), and fluorescence due to GFP (a) was visualized using a microscope equipped with a 488-nm excitation and fluorescein
isothiocyanate filter set for GFP, while the antibody staining of Flag–As-rel2 (c) was visualized using a 570-nm excitation and rhodamine
isothiocyanate filter set. DNA (b, d) was stained with DAPI and visualized using a 372-nm excitation and UV filter set. (B) GFP–As-rel1 and Flag–
As-rel2 were transiently coexpressed in HEK293 cells, and fluorescence due to GFP–As-rel1 (a) and the antibody staining of Flag–As-rel2 (b) were

visualized as described above. DNA (d) was stained with DAPI. Note that GFP–As-rel1 merges with Flag–As-rel2 (c). The nuclei are indicated by
arrowheads.
Ó FEBS 2003 Regulation of Rel by alternative splice variant (Eur. J. Biochem. 270) 4463
constant irrespective of the expression of Flag-tagged As-rel2
(Fig. 4C, lower panel, WB). These results indicate that
As-rel2 has a suppressive effect on the activity of As-rel1,
although it can enter the nucleus escorted by As-rel1.
Isolation of an ascidian IjB homologue
It is well known that the inhibitor IjB modulates the
functions of Rel/NF-jB family members. We have been
trying to isolate cDNA clones encoding ankyrin-repeat
proteins from the ascidian H. roretzi [19], but our attempts
to isolate a cDNA clone encoding an IjB homologue have
not been successful. We therefore searched the EST
database of the ascidian C. intestinalis (Ciona cDNA
resources of Kyoto University) for an IjB homologue.
We found in this database an EST clone, cieg29i06, which is
predicted to include the full length of IjB-like cDNA, and
the nucleotide sequence was determined. The cDNA clone,
designated Ci-IjB, consists of 1041 nucleotides with a
poly(A)-rich tail (Fig. 5A), and its single ORF encodes 347
amino acids containing six ankyrin motifs (Fig. 5B) and
two consensus phosphorylation sequences (DSGXXS)
(Fig. 5A). A homology search revealed that Ci-IjBhas
the highest homology (56%), with human IjBa among the
various IjB members hitherto reported.
Interaction of Ci-IjB with As-rel1 and As-rel2
Next, we carried out an experiment to determine whether
Ci-IjB is capable of forming a complex with As-rel1 or
As-rel2. HEK293T cells transiently expressing Flag-tagged

Ci-IjB with or without c-6Myc-tagged As-rel1 and c-6Myc-
tagged As-rel2 were subjected to the first immunoprecipi-
tation using anti-(Flag-tag) Ig. The eluate obtained by
elution with 3 · Flag-peptide was subsequently subjected to
Western blotting with anti-(c-Myc-tag) Ig. Ci-IjBwas
found to be able to interact with As-rel1 or As-rel2 (Fig. 6A,
middle panel, 1st IP). Next, to determine whether the
Ci-IjB-containing immunoprecipitate also contains both
As-rel1 and As-rel2, the above eluate was subjected to the
second immunoprecipitation using antibody against As-rel1,
and the eluate obtained by elution with 0.1
M
glycine
(pH 3.0) was then subjected to Western blotting with anti-
(c-Myc-tag) Ig. Ci-IjB was found to be able to interact with
both As-rel1 and As-rel2 to form a complex (Fig. 6A, lower
panel, 2nd IP). Thus, Ci-IjB forms a complex with either
As-rel1 or As-rel2 and also with both together. In addition,
to confirm the interaction between As-rel2 and Ci-IjB, we
transiently coexpressed Flag-tagged As-rel2 and c-6Myc-
tagged Ci-IjB in HEK293T cells, and the extracts of
transfected cells were subjected to immunoprecipitation
using anti-(Flag-tag) Ig. The eluate obtained by elution with
3 · Flag-peptide was subsequently subjected to Western
blotting with anti-(c-Myc-tag) Ig (Fig. 6B). It was confirmed
that Ci-IjB is capable of interacting with As-rel2.
As-rel2-dependent effect of Ci-IjB
on As-rel1 localization
AsCi-IjB is capable of interacting with As-rel1 and As-rel2,
we next carried out an experiment to determine whether

Ci-IjB affects the localization of As-rel1 in the presence or
absence of As-rel2. We first transiently expressed GFP–As-
rel1 fusion protein together with c-6Myc-tagged Ci-IjBin
HEK293 cells. As expected from the cases of other Rel/NF-
jB family members, As-rel1 was found to be present in the
cytoplasm (Fig. 7a,b) because Ci-IjB binds to As-rel1 to
Fig. 4. Effect of As-rel2 on transcriptional activity of As-rel1. Tran-
scriptional activities of As-rel1 and As-rel2, transiently expressed in
HEK293T cells, were measured by luciferase assay using GAL4–As-rel
fusion protein and the GAL4-binding DNA sequence. The cells were
transfected with expression vectors containing increasing amounts of
pBIND-As-rel1 (0.18, 0.36, 0.54 and 0.72 lg of DNA) (A), increasing
amounts of pBIND-As-rel2 (0.18, 0.36, 0.54 and 0.72 lg) (B), or
pBIND-As-rel1 (0.18 lg) and increasing amounts of pCI-neo-Flag-
As-rel2 (0.18, 0.36, and 0.54 lg) (C), together with the pG5/luc
reporter vector. Total amounts of transfected DNA were kept constant
(1.5 lg) by adding an empty vector (pBIND or pCI-neo-Flag vector).
The level of activity was normalized on the basis of the level of activity
of control Renilla luciferase. Results are expressed as n-fold induction
in luciferase activity relative to control cells that had been transfected
with an empty vector, pBIND. Triplicate experiments were carried out,
and the error bars represent SD. To assess the expression levels of
GAL4–As-rel1 and Flag–As-rel2, parts of transfected cells were sub-
jected to SDS/PAGE followed by Western blotting (WB) with anti-
(As-rel1) and anti-(Flag-tag) Igs (C, lower panel, WB).
4464 N. Kawai et al.(Eur. J. Biochem. 270) Ó FEBS 2003
sequester it in the cytoplasm. This result is in contrast with
the result obtained from a single transfection with GFP–
As-rel1 fusion protein (Fig. 3A). On the other hand, when
GFP–As-rel1 and c-6Myc-tagged Ci-IjB were transiently

expressed together with Flag-tagged As-rel2 in HEK293
cells, As-rel1 was found to be present in the nucleus, whereas
As-rel2 was located in the cytoplasm (Fig. 7c,d,e). This
situation may arise because As-rel2 sequesters Ci-IjBinthe
cytoplasm to allow As-rel1 to move into the nucleus. Thus,
As-rel2 regulates the localization of As-rel1 in the presence
of Ci-IjB.
As-rel2-dependent effect of Ci-IjB on transcriptional
activity of As-rel1
AsCi-IjB can regulate As-rel1 localization, we next
investigated the effect of Ci-IjB on the transcriptional
activity of As-rel1. We employed the luciferase assay
described above. In HEK293T cells transiently expressing
GAL4–As-rel1 together with Flag-tagged Ci-IjB, luciferase
activity of As-rel1 was inhibited, depending on the dose of
Ci-IjB (Fig. 8A, upper panel, luciferase assay), as expected
from the results on localization shown in Fig. 7. On the
other hand, in the cells transiently expressing GAL4–As-rel1
and Flag-tagged Ci-IjB together with Flag-tagged As-rel2,
the activity of As-rel1 that had been inhibited by Ci-IjBwas
rescued by As-rel2, although to a moderate level (Fig. 8B,
upper panel, luciferase assay). This finding is consistent with
the results in Fig. 7 showing that As-rel1 can move into the
nucleus even in the presence of Ci-IjB when As-rel2 is
expressed with it. It should be noted that the expression
levels of GAL4–As-rel1 remained almost constant irrespec-
tive of the expression of Flag-tagged Ci-IjB and Flag-
tagged As-rel2 (Fig. 8A,B, lower panels, WB). Thus, As-rel2
regulates the transcriptional activity of As-rel1 in the
presence of Ci-IjB.

Discussion
In this study, we first determined that As-rel1 and As-rel2
are splice variants. Next, we demonstrated that As-rel2, a
short splice variant, modulates the localization and tran-
scriptional activity of As-rel1, a typical Rel/NF-jB family
member. In the absence of Ci-IjB, As-rel2 as a heterodimer
with As-rel1 enters the nucleus and suppresses the activity of
As-rel1, whereas in the presence of Ci-IjB, it binds Ci-IjB
and the sequestration of Ci-IjB in the cytoplasm by As-rel2
allows As-rel1 to enter the nucleus, leading to the promotion
of transcription. This is a novel regulatory mechanism for
the function of a Rel/NF-jB family member mediated by a
short splice variant.
As-rel2 is a novel short splice variant of Rel/NF-jB
family proteins which lacks both the nuclear localization
signal and the C-terminal region, a putative transactivation
domain (Fig. 1A). Dorsal B is an alternative splice variant
of Dorsal and it lacks the nuclear localization signal [20].
Dorsal B mRNA is generated because the intron for Dorsal
is not excised in a manner similar to that in the case of
As-rel2 mRNA generation, but, in contrast with the case of
As-rel2, it functions as an activator for transcription when it
can enter the nucleus because it has a C-terminal trans-
activation domain [20].
As expected from the structures of As-rel1 and As-
rel2, we demonstrated that As-rel1 binds to itself and to
As-rel2 to form a homodimer and a heterodimer,
respectively, and that As-rel1 and As-rel2 are localized
in the nucleus and cytoplasm, respectively, when they are
expressed individually (Figs 2 and 3). With regard to the

nuclear localization of As-rel1, it should be noted that
there is little interaction of mammalian IjB proteins with
ascidian As-rel1 (data not shown), and this enabled us to
use cultured mammalian cells for analysis of the inter-
action between As-rel1 and As-rel2 even in the presence
Fig. 5. Sequence and domain structure of
Ci-IjB. (A) Nucleotide and deduced amino-
acid sequences of Ci-IjB. The polyadenylation
signal sequence is underlined by a solid line.
The consensus phosphorylation sequences
are underlined by dotted lines. (B) Domain
structure of Ci-IjB. Note that Ci-IjBcontains
six ankyrin motifs.
Ó FEBS 2003 Regulation of Rel by alternative splice variant (Eur. J. Biochem. 270) 4465
of mammalian endogenous inhibitor IjB proteins. An
unexpected interesting finding in this study is that As-rel2
can enter the nucleus, escorted by As-rel1, and appar-
ently suppresses the transcriptional activity of As-rel1 to
a moderate level (Figs 3B and 4C). This apparent
suppression can be explained if it is assumed that the
heterodimer formed by As-rel1 and As-rel2 has a lower
level of activity than the As-rel1 homodimer.
WetriedtoisolateacDNAcloneforanIjB
homologue from the ascidian H. roretzi, but our attempts
were not successful. Instead of H. roretzi IjB cDNA, a
cDNA for an IjB called Ci-IjB was isolated from
another ascidian, C. intestinalis.TheCi-IjBproteinwas
demonstrated to interact with As-rel1 and As-rel2 and to
suppress the nuclear transport and transcriptional activity
of As-rel1 (Figs 6, 7a and 8A). Interestingly, we found

that the inhibitory effect of Ci-IjB on As-rel1 is
modulated by As-rel2, a short splice variant. When three
proteins, a typical Rel/NF-jB family member, As-rel1, a
short variant, As-rel2, and its inhibitor, Ci-IjB, were
coexpressed, the short variant in the cytoplasm binds
with the inhibitor, enabling As-rel1 to move into the
nucleus, leading to the promotion of transcription (Figs 7
and 8B). This finding provides evidence of a short splice
variant-mediated regulatory mechanism for the function
of a Rel/NF-jB family member. The activity of As-rel1
was rescued from Ci-IjB inhibition by As-rel2 to a
moderate level, comparable with that of the inhibited
activity of As-rel1, when As-rel1 and As-rel2 were
coexpressed in the absence of Ci-IjB (Figs 4C and 8B).
This apparent coincidence in the levels of activity
suggests that the As-rel1–As-rel2 heterodimer, but not
the As-rel1 homodimer, enters the nucleus in the former
case. This explanation, however, is inconsistent with the
results on the cytoplasmic localization of As-rel2 in the
presence of As-rel1 and Ci-IjB (Fig. 7). This discrepancy
cannot be completely explained. Quantitative measure-
ments of interactions between As-rel1, As-rel2, and Ci-
IjB will define it.
In our previous study, Northern blot analysis revealed
that As-rel1 and As-rel2 mRNAs are expressed during
development in H. roretzi [15]. We also showed that
injection of As-rel1 mRNA interfered with H. roretzi
notochord formation, resulting in a shortened tail with a
reduced number of notochord cells and that H. roretzi
embryos coinjected with As-rel1 and As-rel2 mRNAs

developed normally [15]. The results for the single
overexpression of As-rel1 can be explained by its
Fig. 6. Interaction of Ci-IjB with As-rel1 and As-rel2. (A) HEK293T
cells were transiently transfected with the indicated combinations of
c-6Myc-As-rel1, c-6Myc-As-rel2, and Flag-Ci-IjB expression plas-
mids, and 48 h after transfection the cell lysates were subjected to the
first immunoprecipitation (1st IP) using anti-(Flag-tag) Ig. The
immunoprecipitates produced were eluted with 3 · Flag-peptide. The
eluate thus obtained was subjected to the second immunoprecipitation
(2nd IP) using anti-(As-rel1) Ig. The immunoprecipitates produced
were eluted with 0.1
M
glycine (pH 3.0). The eluates obtained from the
first and second immunoprecipitations were subjected to Western
blotting with anti-(c-Myc-tag) and anti-(Flag-tag) Igs. The bands of
As-rel1 and As-rel2 are indicated by a closed arrowhead and an open
arrowhead, respectively. (B) HEK293T cells were transiently trans-
fected with the indicated combinations of Flag-As-rel2 and c-6Myc-Ci-
IjB expression plasmids, and 48 h after transfection the cell lysates
were subjected to immunoprecipitation (IP) using anti-(Flag-tag) Ig.
The immunoprecipitates produced were eluted with 3 · Flag-peptide
and the eluate thus obtained was subjected to Western blotting with
anti-(c-Myc-tag) and anti-(Flag-tag) Igs.
Fig. 7. Localization of As-rel1 and As-rel2 in the presence of Ci-IjB.
HEK293 cells were transiently transfected with GFP-As-rel1 and
c-6Myc-Ci-IjB expression plasmids with (c, d, e) or without Flag-
As-rel2 expression plasmid (a, b). Fluorescence due to GFP-As-rel1
(a, c), DAPI staining of DNA (b, e), and the antibody staining of
Flag-As-rel2 (d) were visualized as described in Fig. 3. The nuclei are
indicated by arrowheads.

4466 N. Kawai et al.(Eur. J. Biochem. 270) Ó FEBS 2003
promotion of target genes that are not expressed
normally, leading to a defect in development, and the
antagonistic effect of As-rel2 on As-rel1 can be explained
by the results of the present study showing that As-rel2
has a suppressive effect on the transcriptional activity of
As-rel1. An H. roretzi IjB homologue has not been
isolated, but Ci-IjB is expressed in embryos of another
ascidian, C. intestinalis (data not shown). If an IjB
homologue is expressed in H. roretzi embryos, As-rel2, in
the presence of the IjB protein, will modulate the
function of As-rel1 in the same manner as that found in
this study. The isolation of an H. roretzi IjB homologue
will lead to clarification of the relationship between As-
rel1 and As-rel2 and their functions in the development
of H. roretzi.
In conclusion, we propose the following novel regula-
tory mechanism of As-rel1 mediated by As-rel2: (a) As-
rel2, in complex with As-rel1, enters the nucleus and
suppresses the transcriptional activity of As-rel1; (b) As-
rel2 binds to IjB in the cytoplasm, resulting in an
increase in the nuclear translocation and transcriptional
activity of As-rel1.
Acknowledgements
We thank Dr Hiroki Takahashi of the National Institute for Basic
Biology for helpful discussion. This work was supported in part by
grants-in-aid from the Ministry of Education, Science, Sports, Culture,
and Technology of Japan.
References
1. Verma, I.M., Stevenson, J.K., Schwarz, E.M., Van

Antwerp, D. & Miyamoto, S. (1995) Rel/NF-jB/IjBfamily:
intimate tales of association and dissociation. Genes Dev. 9,
2723–2735.
2. Ghosh, S., May, M.J. & Kopp, E.B. (1998) NF-jBandRelpro-
teins: evolutionarily conserved mediators of immune responses.
Annu. Rev. Immunol. 16, 225–260.
3. May, M.J. & Ghosh, S. (1998) Signal transduction through NF-
jB. Immunol. Today 19, 80–88.
4. Perkins, N.D. (2000) The Rel/NF-jB family: friend and foe.
Trends Biochem. Sci. 25, 434–440.
5. Karin, M. & Ben-Neriah, Y. (2000) Phosphorylation meets ubi-
quitination: the control of NF-jB activity. Annu. Rev. Immunol.
18, 621–663.
6. Silverman, N. & Maniatis, T. (2001) NF-jB signaling pathway
in mammalian and insect innnate immunology. Genes Dev. 15,
2321–2342.
7. Ghosh, S. & Karin, M. (2002) Missing pieces in the NF-jBpuzzle.
Cell 109, S81–S96.
8. Karin, M. & Lin, A. (2002) NF-jB at the crossroads of life and
death. Nat. Immunol. 3, 221–227.
9. Schmitz, M.L., Bacher, S. & Kracht, M. (2001) IjB-independent
control of NF-jB activity by modulatory phosphorylations.
Trends Biochem. Sci. 26, 186–190.
10. Zhong, H., SuYang, H., Erdjument-Bromage, H., Tempst, P. &
Ghosh, S. (1997) The transcriptional activity of NF-jBisregu-
lated by the IjB-associated PKAc subunit through a cyclic AMP-
independent mechanism. Cell 89, 413–424.
11. Wang, D., Westerheide, S.D., Hanson, J.L. & Baldwin, A.S. Jr
(2000) Tumor necrosis factor a-induced phosphorylation of RelA/
p65 on Ser

529
is controlled by casein kinase II. J. Biol. Chem. 275,
32592–32597.
12. Bhaskar, V., Smith, M. & Courey, A.J. (2002) Conjugation of
Smt3 to dorsal may potentiate the Drosophila immune response.
Mol. Cell. Biol. 22, 492–504.
13. Zhong, H., May, M.J., Jimi, E. & Ghosh, S. (2002) The phos-
phorylation status of nuclear NF-jB determines its association
with CBP/p300 or HDAC-1. Mol. Cell 9, 625–636.
14. Levkau, B., Scatena, M., Giachelli, C.M., Ross, R. & Raines,
E.W. (1999) Apoptosis overrides survival signals through a
caspase-mediated dominant-negative NF-jB loop. Nat. Cell Biol.
1, 227–233.
Fig. 8. Effect of Ci-IjB on transcriptional activity of As-rel1. Tran-
scriptional activity of As-rel1 in the presence of Ci-IjB and/or As-rel2
was measured by luciferase assay as described in Fig. 4. HEK293T
cells were transfected with expression vectors containing pBIND-
As-rel1 (0.05 lg) and increasing amounts of pCI-neo-Flag-Ci-IjB
(0.05, 0.1, and 0.15 lg) (A) or pBIND-As-rel1 (0.05 lg), Flag-As-rel2
(0.05 lg) and increasing amounts of pCI-neo-Flag-Ci-IjB (0.05, 0.1,
and 0.15 lg) (B), together with the pG5/luc reporter vector. Total
amounts of transfected DNA were kept constant (0.5 lg) by adding an
empty vector (pCI-neo-Flag vector). The level of activity was nor-
malized, and results are expressed as fold induction in luciferase
activity as in Fig. 4. Triplicate experiments were carried out, and the
error bars represent SD. To assess the expression levels of GAL4–
As-rel1, Flag–Ci-IjB, and Flag–As-rel2, parts of transfected cells were
subjected to SDS/PAGE followed by Western blotting (WB) with anti-
As-rel1 and anti-(Flag-tag) Igs (A, B, lower panels, WB).
Ó FEBS 2003 Regulation of Rel by alternative splice variant (Eur. J. Biochem. 270) 4467

15. Shimada, M., Satoh, N. & Yokosawa, H. (2001) Involvement of
Rel/NF-jB in regulation of ascidian notochord formation. Dev.
Growth Differ. 43, 145–154.
16. Tran, K., Merika, M. & Thanos, D. (1997) Distinct func-
tional properties of IjBa and IjBb. Mol. Cell. Biol. 17, 5386–5399.
17. Burset, M., Seledtsov, I.A. & Solovyev, V.V. (2000) Analysis of
canonical and non-canonical splice sites in mammalian genomes.
Nucleic Acids Res. 28, 4364–4375.
18. Burset, M., Seledtsov, I.A. & Solovyev, V.V. (2001) SpliceDB:
Database of canonical and non-canonical mammalian splice sites.
Nucleic Acids Res. 29, 255–259.
19. Kondoh, M., Kasai, T., Shimada, M., Kashiwayanagi, M. &
Yokosawa, H. (2003) cDNA cloning and characterization of an
osmotically sensitive TRP channel from ascidian eggs. Comp.
Biochem. Physiol. B132, 417–423.
20.Gross,I.,Georgel,P.,Oertel-Buchheit,P.,Schnarr,M.&
Reichhart, J.M. (1999) Dorsal-B, a splice variant of the Drosophila
factor Dorsal, is a novel Rel/NF-jB transcriptional activator.
Gene 228, 233–242.
4468 N. Kawai et al.(Eur. J. Biochem. 270) Ó FEBS 2003