Proc. Natl. Acad. Sci. USA
Vol. 93, pp. 8241-8246, August 1996
I-TRAF is a novel TRAF-interacting protein that regulates
TRAF-mediated signal transduction
(tumor necrosis factor/tumor necrosis factor receptors/NF-cB/TRAF2)
MIKE ROTHE*, JESSIE XIONG*, HONG-BING SHU*, KEITH WILLIAMSON*, AUDREY GODDARDt,
AND DAVID V. GOEDDEL*t
*Tularik, Inc., Two Corporate Drive, South San Francisco, CA 94080; and tDepartment of Molecular Biology, Genentech, Inc., 460 Point San Bruno Boulevard,
South San Francisco, CA 94080
Contributed by David V Goeddel, May 6, 1996
associated factor (TRAF) proteins associate with and trans-
duce signals from TNF receptor 2, CD40, and presumably
other members of the TNF receptor superfamily. TRAF2 is
required for CD40- and TNF-mediated activation of the tran-
scription factor NF-#cB. Here we describe the isolation and
characterization of a novel TRAF-interacting protein, I-TRAF,
that binds to the conserved TRAF-C domain of the three
known TRAFs. Overexpression of I-TRAF inhibits TRAF2-
mediated NF-KB activation signaled by CD40 and both TNF
receptors. Thus, I-TRAF appears as a natural regulator ofTRAF
function that may act by maintaining TRAFs in a latent state.
Tumor necrosis factor (TNF) receptor-
The tumor necrosis factor (TNF) receptor-associated factor
(TRAF) family of proteins is involved in transducing signals
from various members of the TNF receptor (TNF-R) super-
family (1). TRAF1 and TRAF2 were originally purified as
TNF-R2-associated proteins of45 and 56 kDa, respectively (2).
TRAF3 was identified by two-hybrid interaction cloning as a
CD40-associating protein of 64 kDa (3-6).
The three known members of the TRAF family are com-
posed of distinct structural domains. TRAF1 and TRAF2
share a conserved C-terminal "TRAF domain" of -230 aa
that is involved in homo- and heterooligomerization and
receptor association (2). Inspection of this region in TRAF3
identified two subdomains, the TRAF-C domain, comprising
the C-terminal 150 aa, and the TRAF-N domain, which
consists of a putative coiled-coil structure (4). Whereas the
TRAF-C domain is highly conserved among the three known
TRAFs, the TRAF-N domain is diverged in TRAF3. In
addition, TRAF2 and TRAF3 contain an N-terminal RING
finger and five zinc finger structures of weak sequence simi-
TRAF1 and TRAF2 exist in a multimeric complex that
interacts via TRAF2 with the signaling domains of both
TNF-R2 and CD40 (2, 8). In addition, TRAF2 was recently
found to associate with the TNF-R1-associated death domain
protein (TRADD; ref. 9), a TNF-R1-interacting signaling
protein (10). TRADD can simultaneously bind TNF-R1 and
TRAF2 via distinct domains, thereby recruiting TRAF2 to the
TNF-R1 signaling complex (9).
The two TNF receptors and CD40 can independently gen-
erate signals that lead to the activation of the transcription
factor NF-KB (2, 11-14). Functional analysis demonstrated
that TRAF2 is a common signal transducer for TNF-R2 and
CD40 that mediates activation of NF-KB (8). This effector
function of TRAF2 requires its N-terminal RING finger
domain (8). TRAF2 also appears to be involved in NF-KB
activation by TNF-R1, because overexpression of a dominant
negative version of TRAF2 abolished NF-KB activation sig-
naled by this receptor (9). The role of TRAF3 in signal
transduction is less well defined, but it has been implicated in
CD40-mediated induction of CD23 (4).
Another family of TRAF-interacting proteins was recently
identified through the purification of additional proteins that
associate with TNF-R2 (15). c-IAP1 and c-IAP2 are closely
related cellular members of the inhibitor of apoptosis protein
(IAP) family originally characterized in baculoviruses (16).
The viral IAPs and c-IAPs contain N-terminal baculovirus
IAP repeat motifs and a C-terminal RING finger. The c-IAPs
do not directly contact TNF-R2 but rather associate with
TRAF1 and TRAF2 through their N-terminal baculovirus
IAP repeat motif-comprising domain. The recruitment of
c-IAP1 or c-IAP2 to the TNF-R2 signaling complex requires
a TRAF2-TRAF1 heterocomplex (15). Neither c-IAP1 nor
c-IAP2 is involved in activation of NF-KB by TRAF2 (15).
In an effort to understand how TRAF2 signals downstream
responses such as NF-KB, we initiated a search for TRAF2-
interacting proteins. We isolated I-TRAF, a novel protein that
binds to the TRAF-C domain of the three known TRAFs.
When overexpressed, I-TRAF prevents association ofTRAF2
with TNF-R2 and inhibits activation of NF-KB induced by
TNF-R2 and CD40. Our analysis suggests that I-TRAF reg-
ulates TRAF function by maintaining TRAFs in a latent state.
MATERIALS AND METHODS
Cell Culture and Biological Reagents. Human embryonic
kidney 293 cells were maintained as described (17). Polyclonal
antibody against the N-terminal portion ofhuman TRAF2 was
raised against a 34-mer peptide (HEGIYEEGISILESSSAFP-
DNAARREVESLPAVK) by BabCo (Richmond, CA). The
rabbit anti-I-TRAF antiserum was raised against a glutathione
S-transferase (GST)-I-TRAF fusion protein. The monoclonal
antibody 9E10 against the Myc epitope was provided by R.
Schreiber (Washington University). The anti-FLAG epitope
monoclonal antibodyM2was purchased from Eastman Kodak.
Yeast Two-Hybrid Cloning. Yeast two-hybrid cloning using
full-length TRAF2 in the vector pPC97 as bait (2) was
performed following the Matchmaker Two-Hybrid System
Protocol (CLONTECH). The cDNA libraries screened were
prepared from murine fetal liver stromal cell line 7-4 RNA (2)
and murine peripheral lymph node RNA (provided by P.
Young, D. Dowbenko, and L. Lasky, Genentech). Subsequent
two-hybrid interaction analysis between bait- and prey-
Abbreviations: TNF, tumor necrosis factor; TNF-R, TNF receptor;
TRAF, TNF-R-associated factor; TRADD, TNF-R1-associated death
domain protein; IAP, inhibitor of apoptosis protein; c-IAP, cellular
IAP; I-TRAF, TRAF-interacting protein; GST, glutathione S-
transferase; IL-1, interleukin 1.
Data deposition: The sequences reported in this paper have been
deposited in the GenBank data base (accession nos. U59863 and
4To whom reprint requests should be addressed.
The publication costs of this article were defrayed in part by page charge
payment. This article must therefore be hereby marked "advertisement" in
accordance with 18 U.S.C. §1734 solely to indicate this fact.
Proc. Natl. Acad. Sci. USA 93 (1996)
encoded fusion proteins was carried out in Saccharomyces
cerevisiae Y190 or HF7c cells as described (2, 8, 15).
cDNA Cloning. The cDNA insert from a murine I-TRAF
two-hybrid clone was used as a probe to screen murine CT6
and human HeLa cDNA libraries (2, 10) under standard high
and reduced stringency conditions, respectively (18). The
cDNA inserts of positive phage clones were subcloned into
pBluescript KS (Stratagene) and sequenced on both strands
with Sequenase (United States Biochemical).
Transfections, Reporter Gene Assays, and Electrophoretic
Mobility Shift Assay. For expression in mammalian cells,
cDNAs encoding full-length human I-TRAFa and murine
I-TRAFy were cloned into pRK5 under the transcriptional
control of the cytomegalovirus immediate-early promotor-
enhancer (19). DNA fragments encoding truncated variants of
human I-TRAFa were amplified by PCR. Transient transfec-
tion of 293 cells and reporter gene assays were performed as
described (10, 15).
Induction of NF-KB DNA-binding activity was analyzed by
electrophoretic mobility shift assay, with nuclear extracts
prepared from TNF-stimulated or unstimulated 293 cells as
described (2, 8).
Immunoprecipitation and Immunoblotting. Lysates from
transiently transfected 293 cells were prepared and immuno-
precipitated with anti-FLAG epitope monoclonal antibody or
mouse IgG as control as described (15). Bound proteins were
fractionated by SDS/8% PAGE and transferred to a nitrocel-
lulose membrane. Immunoblot analysis was performed with
rabbit polyclonal anti-I-TRAF antiserum as described (15).
For reimmunoprecipitation analysis, 293 cells were meta-
bolically labeled with [35S]cysteine and [35S]methionine as
described (2), and cell lysates incubated with 10 ,tl of poly-
clonal anti-serum and 40Alof protein G beads. The initial
immune complexwas dissociated byboiling for 5 min in 50mM
Tris-HCl, pH 7.9/0.5% SDS, diluted 20-fold in RIPA buffer
(50 mM Tris-HCl, pH 7.5/150 mM NaCl/1% Nonidet P-40/
0.5% sodium deoxycholate/0.1% SDS), and subjected to a
second round of immunoprecipitation. Bound proteins were
analyzed by SDS/10% PAGE and autoradiography.
Coprecipitation experiments with GST fusion proteins were
performed essentially as described (15). Briefly, aliquots of
lysates from transfected 293 cells (0.5 ml) were incubated for
15 min at 4°C with 10 ,ul of GST-TNF-R2 fusion protein (2)
bound to glutathione agarose beads. The beads were washed
and bound proteins analyzed by SDS/PAGE and immuno-
blotting using anti-FLAG monoclonal antibody. In all cases,
expression of transfected constructs was verified by immuno-
blotting of aliquots of cell lysates.
Generation of GST Fusion Proteins and in Vitro Binding
Assays. I-TRAF was expressed as a GST fusion protein using
the pGEX3X vector (Pharmacia) and purified as described
(20). 35S-labeled proteins were generated with the TNT-
coupled reticulocyte lysate system (Promega) and the various
cDNAs cloned in pBluescript KS (Stratagene) or pRK5. For
each binding assay, 0.5 ,ug of GST-I-TRAF or 0.5Agof GST
bound to glutathione Sepharose beads was incubated with
equivalent cpm of the individual 35S-labeled proteins in 1 ml of
binding buffer (see above) at 4°C for 1 h. Beads were washed
6 times with binding buffer and precipitates were fractionated
by SDS/10% PAGE. The gel was dried and exposed to Kodak
Identification of I-TRAF as a TRAF2-Interacting Protein.
Previous analysis had implicated TRAF2 in TNF-R2- and
CD40-mediated activation of NF-KB (8). To identify potential
downstream components of the TNF-R2/CD40-TRAF2 sig-
naling pathway, we used the yeast two-hybrid system (21) to
screen cDNA libraries for TRAF2-interacting proteins. Mul-
tiple cDNA clones encoding several distinct proteins were
obtained from fetal liver stromal cell and peripheral lymph
node cDNA libraries. Restriction mapping of positive clones
indicated that most were derived from the same gene. Four
fetal liver and two peripheral lymph node cDNA clones were
sequenced. The 5'-ends of six additional murine cDNA clones
isolated by screening a CT6 cDNA library were also sequenced.
Analysis of the obtained DNA sequences revealed cDNAs
corresponding to several distinct transcripts of a murine gene
that was designated TRAF-interacting protein (1-TRAF). Due
to alternative splicing, these transcripts have the potential to
use two different translation initiation codons. The two major
forms of murine I-TRAF mRNA are predicted to encode pro-
teins of 413 and 447 aa that we have termed I-TRAFa and
I-TRAF,3, respectively (Fig. 1). In addition, several splicevariants
resulting in premature termination of I-TRAFa were identified.
The shortest ofthese truncated proteins, I-TRAFy, comprises the
amino terminal 202 amino acids of I-TRAFa (Fig. 1).
Using a murine I-TRAF hybridization probe, we also iso-
lated several human I-TRAF cDNA clones from a HUVEC
cDNA library, all of which encode a 425-aa protein with a
predicted molecular weight of 48,000 that is 82% identical to
murine I-TRAFa (Fig. 1). Data base searches failed to reveal
any proteins having significant sequence similarity to I-TRAF.
Northem blot analysis usingmouse tissues indicated that the -2.4
kb I-TRAF mRNA is ubiquitously expressed (data not shown).
I-TRAF Interacts with the TRAF-C Domains of TRAFI,
TRAF2, and TRAF3. I-TRAF was identified in a yeast two-
hybrid screening using full-length TRAF2 as a bait. To delin-
eate a region in TRAF2 that is required for I-TRAF binding,
we examined the interaction of I-TRAF with various trunca-
tion mutants of TRAF2 in two-hybrid assays. N-terminal
deletion mutants of TRAF2 lacking the RING finger or both
NIP]CIATETQCSVPIQCTDK1 KQEALF PQAKDDINR
Homology between murine and human I-TRAF. An
optimized alignment of the protein sequences of murine and human
I-TRAF is shown. Identical amino acids are boxed. The translation
initiation codons of the murine I-TRAFa and I-TRAF,B splice variants
are indicated by a and 13, respectively. The splice junction is marked
by an asterisk (*). All of the isolated human I-TRAF cDNA clones
correspond to murine I-TRAFa. The initiator methionine of human
I-TRAFa is preceded by an upstream in-frame stop codon not present
in the murine I-TRAFa cDNA clones. The amino acid sequences of
two additional alternative splice variants of mouse I-TRAFa, ending
in premature termination codons, are also listed. The splice junction
of the shortest splice variant, murine I-TRAF-y, is indicated by y.
Biochemistry:Rothe et al.
Proc. Natl. Acad. Sci. USA 93 (1996)8243
the RING finger and the five zinc finger structures were still
able to bind I-TRAF (Table 1). No interaction could be
detected between I-TRAF and a C-terminal deletion mutant
of TRAF2 lacking the TRAF-C domain (Table 1). These
findings indicate that I-TRAF binds to the conserved TRAF-C
domain (amino acids 359-501) ofTRAF2. Furthermore, both
TRAF1 and TRAF3 also associated with I-TRAF (Table 1).
The interaction ofI-TRAF with TRAF1, TRAF2, andTRAF3
was confirmed by binding experiments in which human I-
TRAF expressed as a GST fusion protein was found to interact
specifically with 35S-labeled TRAF proteins (Fig. 2A). Results
from both two-hybrid and in vitro binding experiments showed
that I-TRAF interacts more strongly withTRAF1 andTRAF2
than with TRAF3 (Table 1; Fig. 2A).
Inspection of the various I-TRAF cDNAs obtained by
two-hybrid screening indicated that I-TRAFy, consisting of
the N-terminal portion of murine I-TRAF (amino acids 35-
236), is sufficient for interaction with TRAF2 (Table 1). To
further investigate the interaction ofTRAF2 with I-TRAF, we
used a transfection-based coimmunoprecipitation assay. An
expression vector encoding FLAG epitope-tagged TRAF2was
transfected alone or with expression vectors encoding full-
length or deletion mutants ofhuman I-TRAFa into embryonic
kidney 293 cells. Cell lysates were immunoprecipitated using a
monoclonal antibody against the FLAG epitope, and copre-
cipitating I-TRAF was detected by immunoblotting with poly-
clonal anti-I-TRAF antibodies. Consistent with the results
obtained by two-hybrid analysis, TRAF2 coprecipitated both
full-length I-TRAFa and mutant I-TRAFa comprising the
N-terminal 212 aa [I-TRAFa(1-212)] (Fig. 2B). Furthermore,
TRAF2 was also able to specifically coprecipitate the C-
terminal half of human I-TRAFa [I-TRAFa(213-425)] (Fig.
2B). These findings suggest that multiple regions within I-
TRAF may mediate its association with TRAF2.
Endogenous Association of I-TRAF and TRAF2 in Mam-
malian Cells. To confirm the interaction of I-TRAF and
TRAF2 in a native system endogenous TRAF2 was immuno-
precipitated from lysates ofuntreated or TNF-treated 293 cells
Interactions between I-TRAF and TRAFs
Yeast Y190 cells were cotransformed with expression vectors en-
coding Gal4 activation domain-I-TRAF fusion proteins and Gal4
DNA-binding domain expression vectors as indicated. Each transfor-
mation mixture was plated on a synthetic dextrose plate lacking leucine
and tryptophan. Filter assays for ,3-galactosidase activity were per-
formed to detect interaction between fusion proteins. Double plus and
plus signs indicate strong blue color development within 30 min and
1 h of the assay, respectively. Minus sign indicates no development of
color within 24 h. "Three-hybrid" interaction analysis was performed
in yeast HF7c cells as described (2).
*Plus and minus signs indicate growth or lack of growth, respectively,
of transformed yeast colonies on plates lacking tryptophan, leucine
and histidine. Control transformations with empty Gal4 vectors were
negative and are not listed.
I-TRAF with TRAF1, TRAF2, and TRAF3. The interactions of 35S-
labeled TRAF1, TRAF2, TRAF3, TNF-R2, and TRADD with GST-I-
TRAF and control GST protein were examined as described. (B)
Interaction of TRAF2 with I-TRAF mutants. 293 cells were transiently
transfected with an expression vector (2.5 ,g) encoding FLAG epitope-
taggedTRAF2 and I-TRAFa expression constructs (2.5 ,ug), as indicated
by plus signs. After 24 h, cell extracts were immunoprecipitated with
anti-FLAG monoclonal antibody (aFLAG) or control IgG (IgG). Co-
precipitating I-TRAF was detected by immunoblotting with polyclonal
anti-I-TRAF antibodies. (C) Endogenous association of I-TRAF and
TRAF2 in mammalian cells. 293 cells were metabolically labeled with
[35S]cysteine and [35S]methionine and left untreated (lanes 1 and 2) or
stimulated with human TNF (100 ng/ml) for 15 min (lanes 3 and 4). Cell
lysates were immunoprecipitated with anti-TRAF2 antibodies (lanes
1-4), and the initial immune complex was dissociated and subjected to a
second round of immunoprecipitation with anti-I-TRAF antibodies
(lanes 2 and 4) or preimmune serum (lanes 1 and 3). The position of
coprecipitated I-TRAF is marked by an arrow. Positions of molecular
mass standards (in kDa) are shown on the left in all panels.
Interaction of I-TRAF with TRAFs. (A) Association of
thatwere metabolically labeled with [35S]cysteine and [35S]me-
thionine. The immune complex was dissociated and subjected
to a second round of immunoprecipitation with anti-I-TRAF
Biochemistry:Rothe et aL
Proc. Natl. Acad. Sci. USA 93 (1996)
antibodies or preimmune serum as control. The anti-I-TRAF
antibodies specifically precipitated a labeled protein of -48
kDa (Fig. 2C), corresponding in size to overexpressed I-TRAF
(see above). In 293 cells, this association of I-TRAF and
TRAF2 was observed independent of TNF stimulation (Fig.
2C; see below).
I-TRAF and c-IAP Bind to Distinct Sites Within the TRAF
Domain ofTRAF2. Next we examined the interaction between
I-TRAF and the TRAF2-associating protein c-IAP1 (15).
Human I-TRAF and FLAG epitope-tagged c-IAP1 were
transiently coexpressed in 293 cells, cell lysates were immu-
noprecipitated with monoclonal anti-FLAG epitope antibody,
and coprecipitating I-TRAF was detected by immunoblotting
with polyclonal anti-I-TRAF antibodies. In this assay, I-TRAF
did not directly associate with c-IAP1 (Fig. 3). However, when
I-TRAF and c-IAP1 were coexpressed in the presence of
overproduced TRAFM, TRAF2, or both TRAF proteins,
strong coprecipitation of I-TRAF and c-IAP1 was observed
(Fig. 3; data not shown). Thus, I-TRAF and c-IAP1 bind to
nonoverlapping sites within theTRAF domains ofTRAF1 and
TRAF2. This is in agreement with results from two-hybrid
analysis, which mapped the binding of I-TRAF and c-IAP1 to
the TRAF-C and TRAF-N domains of TRAF2, respectively
(see above and ref. 15).
I-TRAF Prevents Association of TRAF2 with TNF-R2. I-
TRAF does not directly interact with TNF-R2 (Table 1; Fig.
2A). However, as TRAF1, TRAF2, and c-IAPs can form a
complex with TNF-R2, it was important to ask if I-TRAF can
indirectly associate with TNF-R2 via TRAFs. A three-hybrid
interaction test was performed to address this question.
Whereas TRAF2 can bind simultaneously to TRAF1 and
TNF-R2 in this assay (2), it was not able to mediate I-TRAF
interaction with TNF-R2 (Table 1). In fact, I-TRAF expres-
sion in yeast was found to inhibit the association of TRAF2
with TNF-R2 (data not shown). This result is consistent with
I-TRAF and TNF-R2 both binding to the TRAF-C domain of
TRAF2 (see above and ref. 15).
The observed inhibitory effect of I-TRAF on the TRAF2-
TNF-R2 interaction was further investigated in mammalian
cells. TRAF2 containing an N-terminal FLAG epitope was
expressed in human 293 cells. When cell extracts were incu-
bated with a fusion protein of GST and the cytoplasmic
domain of TNF-R2 (GST-TNF-R2), specific binding of
TRAF2 was detected using anti-FLAG monoclonal antibody
(Fig. 4). Coexpression of TRAF2 with increasing amounts of
siently transfected with I-TRAFa, TRAF2, TRAFM
epitope-tagged c-IAP1 expression vectors (2.5 ,ug) as indicated by plus
signs. After 36 h, cell lysates were immunoprecipitated with anti-
FLAG monoclonal antibody. Coprecipitating I-TRAFa was detected
by immunoblot analysis using polyclonal anti-I-TRAF antibodies. The
positions of molecular mass markers (in kDa) are shown on the left.
Interaction of I-TRAF with c-IAP1. 293 cells were tran-
cells were transiently transfected with the indicated amounts (in /-g)
of I-TRAFa and FLAG epitope-tagged TRAF2 expression vectors.
After 24 h, cell lysates were incubated with GST-TNF-R2 fusion
protein bound to glutathione agarose beads as described. Coprecipi-
tating TRAF2 was detected by immunoblotting with monoclonal
anti-FLAG antibody. The interaction of TRAF2 (arrow) with the
GST-TNF-R2 fusionproteinis inhibitedbyincreased coexpressionof
I-TRAF. Thepositionsof molecular mass markers (in kDa) are shown
on the left.
I-TRAF blocks association of TRAF2 with TNF-R2. 293
I-TRAF effectively blocked the TRAF2-TNF-R2 interaction
(Fig. 4). Furthermore, no I-TRAF was precipitated by the
TRAF2-GST-TNF-R2 complex (data not shown). Nor could
I-TRAF be coimmunoprecipitated from 293 cells with the
TNF-R2-TRAF1/2 complex (data not shown). These results
show that TRAF2 is not able to bind TNF-R2 and I-TRAF
simultaneously, possibly due to equivalent or overlapping
binding sites in the TRAF-C domain. Since 293 cells do not
possess endogenous TNF-R2 capable of recruiting TRAF2
into the receptor signaling complex (8, 9), the observed
association ofTRAF2 and I-TRAF in these cells is constitutive
(see above; Fig. 2C.
I-TRAF Overexpression Inhibits TRAF2-Mediated NF-RcB
Activation. The ability of I-TRAF to bind the three known
TRAF family members raised the possibility thati-TRAF
functions as a general regulator of TRAF-mediated signaling
events. Inparticular,the observedinhibitoryeffect of i-TRAF
on TRAF2-TNF-R2 interaction suggested that I-TRAF ex-
pression might negatively influence TRAF2 signal transduc-
tion. Consequently, we measured the effect of I-TRAF ex-
pression on TRAF2-mediated NF-KB activation. Electro-
phoretic mobility shift assay showed that transient expression
of TRAF2 in 293 cells potently induces NF-KB DNA-binding
activity (8),whereas overexpressionof I-TRAFa alone had no
such effect (Fig. 5A).Whenoverexpressed, i-TRAFa dramat-
ically inhibited TRAF2-mediated NF-KB activation (Fig. 5A)
in a manner similar to overexpression of a mutant TRAF
protein (TRAF2(87-501)) that exerts a dominant negative
effect on TRAF2 signaling (ref. 8; Fig. 5A). The inhibitory
effect of I-TRAF on TRAF2-mediated NF-KB activation was
further investigated using a NF-KB-dependent reporter gene
(22). In this assay, overexpression of i-TRAFa inhibited
TRAF2-induced reporter geneactivation in adose-dependent
manner (Fig. SB).
We next examined the effect of i-TRAF overexpression on
NF-KB activation triggered by the TRAF2-interacting recep-
tors CD4O and TNF-R2. Transient expression of these recep-
tors has been shown to induce ligand-independent receptor
aggregation, which activates N-F-KB inae TRAF2-dependent
process (8).As observed above for TRAF2, NF-KB activation
through both TNF-R2 (Fig. SC) and CD40 (Fig. 5D) was
effectively blocked by increased expression of I-TRAFa.
Biochemistry:Rothe et al.
Proc. Natl. Acad. Sci. USA 93 (1996) 8245
I-TRAFa DNA (jg)
activation signaled by TRAF2. (A) Inhibition of Induction of NF-KB
DNA-binding activity by I-TRAF. 293 cells were transiently trans-
fected with pRK control vector (lanes 1 and 2) or expression vectors
for TRAF2 (3 jtg; lanes 5-7), I-TRAF (12 ,ug; lanes 3 and 6), and
TRAF2(87-501) (12 ,ug; lanes 4 and 7). Cells were stimulated with
human TNF (100 ng/ml) (lane 2) for
untreated (lanes 1 and 3-7). Nuclear extracts were prepared 24 h after
transfection and analyzed for NF-KB DNA-binding activity by elec-
trophoretic mobility shift assay with a radiolabeled double-stranded
oligonucleotide containing two NF-kB sites. B refers to oligonucleo-
tide probe in a complex with protein. (B-E) Inhibition of NF-KB-
dependent reporter gene activation by I-TRAF. 293 cells were tran-
siently transfected with an E-selectin promotor-luciferase reporter
gene plasmid (0.5 ,ug; B-E) and increasing amounts (in ,ug) of
expression vectors for I-TRAFa (B-E), in the presence of TRAF2
expression vector (0.5,ug; B) or expression vectors for murine TNF-R2
(0.1 ,ug; C) or CD40 (1,ug; D). Each transfection also contained 1 ,ug
of pRSV-f3gal. After 24 h (B-D), luciferase activities were determined
and normalized on the basis of 13-galactosidase expression. In E, cells
were harvested 42 h after transfection following stimulation for 6 h
with 20 ng/ml of either TNF (closed bars) or IL-1 (hatched bars).
Values relative to control transfections containing the reporter gene
plasmid and empty vectors
representative experiment in which each transfection was performed
Inhibitory effect of I-TRAF overexpression
1 h before harvest
± SEM for one
Effect of I-TRAF on NF-KcB Activation Signaled by TNF-R1
and Interleukin 1 (IL-1). Since TRAF2 is also required for
NF-KB activation signaled by TNF-R1 (9), we investigated the
effect of I-TRAF overproduction on NF-KB-dependent re-
porter gene activation triggered by this receptor. Overexpres-
sion of I-TRAF inhibited TNF-induced NF-KB activation in
293 cells (Fig. SE), which is exclusively mediated through
endogenous TNF-R1 (8). This finding is consistent with the
involvement of TRAF2 in TNF-R1-mediated NF-KB activa-
tion (9) and with our observation that I-TRAF overexpression
can negatively regulate TRAF2 function.
Whereas TRAF2 is required for TNF-induced NF-KB acti-
vation, overexpression of dominant negative TRAF2 has no
effect on NF-KB activation induced by IL-1 in 293 cells (9).
Surprisingly, when expressed at high levels, I-TRAF abolished
IL-1-mediated NF-KB activation in 293 cells (Fig. 5E). Thus,
I-TRAF appears to possess a general ability to inhibit NF-KB
activation triggered by diverse stimuli. To exclude that I-TRAF
nonspecifically suppresses reporter gene activation in 293 cells,
we examined its effect on an interleukin-4-responsive reporter
gene (U. Schindler, personal communication). In this assay,
overexpression of I-TRAF did not inhibit interleukin 4-in-
duced reporter gene activation (data not shown).
The recent identification of distinct classes of receptor-
associated signal transducers provided insights into how mem-
bers of the TNF-R superfamily initiate downstream signaling
events (reviewed in ref.
TRAF domains were initially characterized based on their
direct association with the cytoplasmic domains of TNF-R2
and CD40, whereas death domain-containing proteins were
found to interact with TNF-R1 and the Fas antigen. More
recently, it became apparent that both TRAF domain proteins
as well as death domain proteins function as adaptors that
recruit additional signaling proteins into the cognate receptor
complexes. For example, TRAFs mediate the indirect associ-
ation of TNF-R2 with members of the IAP family, c-IAP1 and
c-IAP2 (15). Also, TRADD interacts with both TRAFs and
other death domain proteins, thereby enabling them to bind to
the TNF-R1 signaling complex (9).
Recognizing the adaptor function of TRAF2, we embarked
on a yeast two-hybrid screen for TRAF2-interacting proteins
to identify candidate downstream components of the TNF-
R2/CD40-TRAF2 signaling pathway. We found that the
cDNAs isolated most frequently encoded a novel protein we
termed I-TRAF. I-TRAF does not interact with the RING
finger or zinc finger domains of TRAF2 that have been
implicated in TRAF2-dependent NF-KB activation (ref. 8; M.
Takeuchi, M.R., and D.V.G., unpublished data), but rather it
binds to the conserved TRAF-C domains of the three known
TRAFs. Based on these findings
I-TRAF is specifically involved in mediating TRAF2-induced
NF-KB activation. This conclusion is supported by the failure
of I-TRAFa overexpression to induce activation of NF-KB in
mammalian cells. Therefore, we conclude that I-TRAF may be
a general regulator of TRAF protein function.
A unique property of I-TRAF that distinguishes it from
TRAF2-interacting proteins such as c-IAPs is that I-TRAF
cannot be recruited into the TNF-R2 signaling complex. This
is because both I-TRAF and TNF-R2 bind to the same or
overlapping sites in the TRAF-C domain of TRAF2. In fact,
I-TRAF was found to prevent association of TRAF2 with
TNF-R2. Similarly, I-TRAF overexpression inhibited TRAF2-
dependent NF-KB activation induced by overexpression of
TRAF2, TNF-R2, and CD40. This inhibitory activity of I-
TRAF localizes in its C-terminal domain (J.X., M.R., and
D.V.G., unpublished results).
1). Signaling proteins containing
it appears unlikely that
Biochemistry:Rothe et al.
Proc. Natl. Acad. Sci. USA 93 (1996)
Our analysis of I-TRAF is consistent with a model in which
I-TRAF functions as a general regulator of TRAF-mediated
signaling events (Fig. 6). In particular, we speculate that
I-TRAF may be a natural inhibitor of TRAF function that
regulates TRAF protein activity by sequestering TRAFs in a
latent state in the cytoplasm. By bindingTRAFs in the absence
of ligand (CD40L or TNF), I-TRAF may prevent continuous
signaling by inhibiting spontaneous TRAF aggregation. This
hypothesis is in agreement with the finding that overexpression
ofTRAF2 in the absence of receptor clustering is sufficient to
trigger NF-KB activation (8). Ligand-induced aggregation of
CD40 or TNF-R2 might then be expected to release TRAFs
from I-TRAF inhibition by providing a new, higher affinity
TRAF binding site. Dissociation of I-TRAF would allow
TRAFs to translocate to the cytoplasmic membrane, where
they initiate specific signaling cascades (Fig. 6). In this model,
various TRAF-interacting signaling proteins such as c-IAPs
are recruited to the TNF-R2 signaling complex by virtue of
their association with TRAFs. Although ligand-dependent
association ofTRAFs with their cognate receptors has not yet
been demonstrated, binding of the death domain proteins
TRADD and FADD to TNF-R1 (9) and the Fas antigen (23),
respectively, has been shown to be induced by ligand stimu-
lation. Similarly, the indirect recruitment of TRAF2 into the
TNF-R1 signaling complex via TRADD occurs in a ligand-
dependent manner (H.-B.S. and D.V.G., unpublished data).
i i-Q r%t: -wt9LUI
/SEr /ATiTA,liT; /;ITEATITITITIT/TITI
/ ! /
FIG. 6. A model for the activation ofthe NF-KB signal transduction
pathway by the TNF-R2-TRAF signaling complex
Alternatively to our proposed model, it is possible that
I-TRAF may act to turn off or reset TRAF2 activation signals
after completion of receptor-mediated signaling. I-TRAF may
also play a role in regulating TRAF protein stability in that its
association with TRAFs could either exert a stabilizing effect
or target TRAFs for degradation. A detailed characterization
of I-TRAF's role in TNF receptor signaling will have to await
targeted disruption of its encoding gene in mice.
We thank Qimin Gu for help with the DNA sequencing, Vishva
Dixit for providing the TRAF3 cDNA, Larry Lasky for the PLN
two-hybrid cDNA library, and Laura Medin for help with the artwork.
We also thank Hailing Hsu for the initial observation of I-TRAF's
inhibitory effect on NF-KB activation by IL-1.
Vandenabeele, P., Declercq, W., Beyaert, R. & Fiers, W. (1995)
Trends Cell Biol. 5, 392-399.
Rothe, M., Wong, S. C., Henzel, W. J. & Goeddel, D. V. (1994)
Cell 78, 681-692.
Hu, H. M., 0'Rourke, K., Boguski, M. S. & Dixit, V. M. (1994)
J. Biol. Chem. 269, 30069-30072.
Cheng, G., Cleary, A. M., Ye, Z.-s., Hong, D., Lederman, S. &
Baltimore, D. (1995) Science 267, 1494-1498.
Mosialos, G., Birkenbach, M., Yalamanchili, R., VanArsdale, T.,
Ware, C. & Kieff, E. (1995) Cell 80, 389-399.
Sato, T., Irie, S. & Reed, J. C. (1995) FEBS Lett. 358, 113-118.
Song, H. Y. & Donner, D. B. (1995) Biochem. J. 809, 825-829.
Rothe, M., Sarma, V., Dixit, V. M. & Goeddel, D. V. (1995)
Science 269, 1424-1427.
Hsu, H., Shu, H.-B., Pan, M.-P. & Goeddel, D. V. (1996) Cell 84,
Hsu, H., Xiong, J. & Goeddel, D. V. (1995) Cell 81, 495-504.
Wiegmann, K., Schutze, S., Kampen, E., Himmler, A., Machleidt,
T. & Kronke, M. (1992) J. Biol. Chem. 267, 17997-18001.
Lwgreid, A., Medvedev, A., Nonstad, U., Bombara, M. P.,
Ranges, G., Sundan, A. & Espevik, T. (1994) J. Biol. Chem. 269,
Berberich, I., Shu, G. L. & Clark, E. A. (1994) J. Immunol. 153,
Sarma, V., Lin, Z., Rust, B. M., Tewari, M., Noelle, R. J. & Dixit,
V. M. (1995) J. Biol. Chem. 270, 12343-12346.
Rothe, M., Pan, M.-G., Henzel, W. J., Ayres, T. M. & Goeddel,
D. V. (1995) Cell 83, 1243-1252.
Clem, R. J. & Miller, L. K. (1994) in Induction and Inhibition of
Apoptosis byInsect Viruses, eds. Tomei, D. L. & Cope, F. 0. (Cold
Spring Harbor Lab. Press, Plainview, NY), pp. 89-110.
Pennica, D., Lam, V. T., Mize, N. K., Weber, R. F., Lewis, M.,
Fendly, B. M., Lipari, M. T. & Goeddel, D. V. (1992) J. Biol.
Chem. 267, 21172-21178.
Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989) Molecular
Cloning: A Laboratory Manual (Cold Spring Harbor Lab. Press,
Schall, T. J., Lewis, M., Koller, K. J., Lee, A., Rice, G. C., Wong,
G. H. W., Gatanaga, T., Granger, G. A., Lentz, R., Raab, H.,
Kohr, W. J. & Goeddel, D. V. (1990) Cell 61, 361-370.
Smith, D. B. & Johnson, K. S. (1988) Gene 67, 31-40.
Fields, S. & Song, O.-k. (1989) Nature (London) 340, 245-246.
Schindler, U. & Baichwal, V. R. (1994) Mol. Cell. Biol. 14,
Kischkel, F. C., Hellbardt, S., Behrmann, I., Germer, M., Pawlita,
M., Krammer, P. H. & Peter, M. E. (1995) EMBO J. 14, 5579-
Biochemistry:Rothe et al.