Acetylation of Stat1 modulates NF-kappaB activity.

Abstract

Acetylation of signaling molecules can lead to apoptosis or differentiation of carcinoma cells. The molecular mechanisms underlying these processes and the biological role of enzymes mediating the transfer or removal of an acetyl-group are currently under intense investigation. Our study shows that Stat1 is an acetylated protein. Stat1 acetylation depends on the balance between Stat1-associated histone deacetylases (HDACs) and histone acetyltransferases (HATs) such as CBP. Remarkably both inhibitors of HDACs and the cytokine interferon alpha alter this equilibrium and induce Stat1 acetylation. The analysis of Stat1 mutants reveals Lys 410 and Lys 413 as acetylation sites. Experiments with Stat1 mutants mimicking either constitutively acetylated or nonacetylated states show that only acetylated Stat1 is able to interact with NF-kappaB p65. As a consequence, p65 DNA binding, nuclear localization, and expression of anti-apoptotic NF-kappaB target genes decrease. These findings show how the acetylation of Stat1 regulates NF-kappaB activity and thus ultimately apoptosis.

Full text

Page 1

Acetylation of Stat1 modulates
NF-?B activity
Oliver H. Krämer,1,4Daniela Baus,1,3Shirley K. Knauer,1,3Stefan Stein,1Elke Jäger,2
Roland H. Stauber,1Manuel Grez,1Edith Pfitzner,1and Thorsten Heinzel1,4,5
1Georg-Speyer-Haus, D-60596 Frankfurt, Germany;2Medizinische Klinik II-Onkologie, Krankenhaus Nordwest,
D-60488 Frankfurt, Germany
Acetylation of signaling molecules can lead to apoptosis or differentiation of carcinoma cells. The molecular
mechanisms underlying these processes and the biological role of enzymes mediating the transfer or removal
of an acetyl-group are currently under intense investigation. Our study shows that Stat1 is an acetylated
protein. Stat1 acetylation depends on the balance between Stat1-associated histone deacetylases (HDACs) and
histone acetyltransferases (HATs) such as CBP. Remarkably both inhibitors of HDACs and the cytokine
interferon ? alter this equilibrium and induce Stat1 acetylation. The analysis of Stat1 mutants reveals Lys 410
and Lys 413 as acetylation sites. Experiments with Stat1 mutants mimicking either constitutively acetylated
or nonacetylated states show that only acetylated Stat1 is able to interact with NF-?B p65. As a consequence,
p65 DNA binding, nuclear localization, and expression of anti-apoptotic NF-?B target genes decrease. These
findings show how the acetylation of Stat1 regulates NF-?B activity and thus ultimately apoptosis.
[Keywords: Stat1; NF-?B; acetylation; histone deacetylase; Interferon ?; HDAC inhibitor]
Received August 26, 2005; revised version accepted December 19, 2005.
Many signal transduction pathways ultimately result in
the post-translational modification of histones, which
determines the expression of genes important for cell
growth, differentiation, and apoptosis (Wolffe and Hayes
1999; Schreiber and Bernstein 2002). Acetylation of the
N-terminal tails of histones correlates with gene activa-
tion, while histone deacetylation mediates transcrip-
tional repression (Strahl and Allis 2000). It has also be-
come clear that regulated acetylation of nonhistone pro-
teins determines cellular fate and survival (Blobel 2000;
Kouzarides 2000; Cohen et al. 2004). The fine-tuned
equilibrium of protein acetylation and deacetylation is
maintained by histone acetyltransferases (HATs) and
histone deacetylases (HDACs) (Kouzarides 1999). This
constitutes a precise and dynamic regulatory system
modulating protein assemblies in response to external
and internal signals. Protein acetylation therefore ap-
pears to play a role that is comparable to phosphoryla-
tion (Schreiber and Bernstein 2002).
HDAC inhibitors (HDACi) have been shown to change
the expression pattern of genes involved in differentia-
tion, cell cycle arrest, and apoptosis, and they are con-
sidered as candidate drugs for cancer therapy (Krämer et
al. 2001; Kelly et al. 2002; Melnick and Licht 2002).
However, the molecular mechanisms underlying cell-
specific modulation of signaling pathways by HDACi
and factors determining sensitivity toward these com-
pounds are still subject to intense investigation (Mayo et
al. 2003). Several recent reports suggest that HDACi-in-
duced apoptosis depends on the expression of Jun, Bcl-2
proteins, p21WAF/CIP1, p53, NF-?B, and Akt (Vrana et al.
1999; Henderson et al. 2003; Mayo et al. 2003). For some
of these proteins the association with HDACs and/or
acetylation of lysine residues has been shown (Kouza-
rides 2000; Chen and Greene 2003; Kiernan et al. 2003;
Weiss et al. 2003).
In the case of NF-?B, which contributes significantly
to anti-apoptotic signaling (Perkins 2004), HDACi were
shown to repress NF-?B signaling and expression of sev-
eral NF-?B target genes (Huang et al. 1997; Inan et al.
2000; Krämer et al. 2001). Given the critical role of NF-
?B in tumorigenesis, several studies were undertaken to
identify factors influencing this tumor promoter (Per-
kins 2004). It became clear that NF-?B signaling is con-
trolled at several levels by regulatory proteins, such as
the I-?B protein family. Furthermore, Stat1 has been sug-
gested to repress NF-?B-mediated signaling (Wang et al.
2000; Suk et al. 2001; Shen and Lentsch 2004), and sev-
eral stimuli induce apoptosis to a significantly greater
extent in a Stat1-positive cellular background (Kumar et
al. 1997; Meyer et al. 2002). Similar to NF-?B, Stat1 as-
3These authors contributed equally to this work.
4Present address: Institute of Biochemistry and Biophysics, Friedrich-
Schiller-University of Jena, D-07743 Jena, Germany
5Corresponding author.
E-MAIL T.Heinzel@uni-jena.de; FAX 49-3641-949352.
Article and publication are at http://www.genesdev.org/cgi/doi/10.1101/
gad.364306.
GENES & DEVELOPMENT 20:473–485 © 2006 by Cold Spring Harbor Laboratory Press ISSN 0890-9369/06; www.genesdev.org473

Page 2

sociates with HATs and HDACs (Korzus et al. 1998;
Nusinzon and Horvath 2003). Stat1 regulates the expres-
sion of gene products mediating various cellular pro-
cesses constitutively or inducibly by growth factors and
cytokines such as interferons in a phosphorylation-de-
pendent manner (Chatterjee-Kishore et al. 2000; Ihle
2001). Induction of interferon signaling has been shown
to activate the apoptotic program, which indicates that
these cytokines could be useful in cancer therapy (de
Vries et al. 2003). Remarkably, tyrosine phosphorylation
of Stat1 and its transcriptional activity appear to be dis-
pensable for NF-?B inhibition and apoptosis induction in
response to certain stimuli (Wang et al. 2000; Meyer et
al. 2002). However, it is still unclear which other post-
translational modifications are involved in this process
and whether conditions exist in which Stat1 and NF-?B
can interact with each other.
We investigated the molecular mechanisms underly-
ing the induction of apoptosis and the modulation of
signaling pathways by HDACi and interferon ? in hu-
man melanoma cell lines. Our results show that cells
undergoing apoptosis in response to such substances in-
crease expression and acetylation of Stat1. This leads to
altered interactions of Stat1 with HDACs, CBP, and NF-
?B. Our results indicate that the acetylation of Stat1
functions as a molecular switch, which permits binding
of Stat1 to NF-?B and thus reduces NF-?B signaling. This
mechanism appears to be critical for the induction of cell
death by pharmacological and physiological stimuli.
Results
Response of human melanoma cells to HDACi
HDACi can induce growth arrest and apoptosis in tumor
cells of different origin (Krämer et al. 2001; Kelly et al.
2002). We found significant differences in the sensitivity
of various melanoma cell lines toward these compounds.
SK-37 cells showed strong growth reduction in the MTT
assay upon HDACi treatment, whereas NW-1539 cells
were not affected significantly at the inhibitor concen-
trations used (Fig. 1A). These cell lines are prototypical
examples and have characteristics similar to other mela-
noma cell lines that are either HDACi sensitive (e.g.,
MZ-19) or resistant (e.g., NW-450).
To investigate the molecular mechanisms underlying
this differential response, we evaluated whether the re-
duced proliferation due to HDACi relies on proapoptotic
properties and effects on caspases (Thornberry and Lazeb-
nik 1998). In the HDACi-sensitive SK-37 cell line, we
detected activation of the initiator caspases 8 and 9 after
Figure 1.
MTT test after exposure to VPA (0.5–5 mM) or TSA (100 nM) for 48 h (SK-37) or 72 h (NW-1539); (0) untreated cells. (B) Induction of
activated caspase 9 (Casp 9, activation denoted by an asterisk) and cleavage of full-length caspase 8 (Casp 8 fl) into the active subunits
p43/41/18 were detected by Western blot after treatment of SK-37 cells with VPA (V, 1.5 mM) or TSA (T, 100 nM) for 48 h. Caspase
3 activity was measured by conversion of Ac-DEVD-pNA to pNA, which has an absorption peak at 405 nm. This increase is given
relative to the activity of lysates from untreated cells (Ctl). HDACi-induced conversion of full-length caspase 3 (Casp 3 fl) to the active
p17/19 subunits was analyzed in SK-37 and NW-450 cells. (C) Proteolytic cleavage of PARP and apoptotic chromatin fragmentation
induced by VPA (1.5 mM) or TSA (100 nM) after 48 h were detected by Western blot and PI FACS analysis. Cotreatment of SK-37 cells
with Z-VAD-FMK (Z, 100 µM) blocks HDACi-induced apoptosis.
HDACi induce apoptosis in SK-37 cells. (A) The proliferation of SK-37 and NW-1539 melanoma cells was determined by
Krämer et al.
474 GENES & DEVELOPMENT

Page 3

treatment (Fig. 1B). Furthermore, we measured caspase 3
activity in extracts from these cells by colorimetric as-
say. Conversion of the proenzyme form of the execu-
tioner caspase 3 (p32) to the catalytically active proteases
p17 and p19 was detectable in SK-37 but not in HDACi-
resistant NW-450 cells (Fig. 1B).
Activation of caspase 3 during HDACi-mediated apo-
ptosis of SK-37 cells was also verified by examining the
cleavage of PARP (116 kDa) into 85- and 28-kDa frag-
ments (Thornberry and Lazebnik 1998). The pan-caspase
inhibitor Z-VAD-FMK inhibited PARP cleavage as well
as the occurrence of a hypodiploid (sub-G1) fraction re-
sulting from DNA fragmentation in SK-37 cells (Fig. 1C).
Analysis of nuclei stained with Hoechst dye gave similar
results (data not shown). These observations confirm
that HDACi trigger apoptotic, caspase-dependent path-
ways in SK-37 melanoma cells (Fig. 1C). Furthermore, no
signs of nonspecific cell permeabilization and necrotic
cell death were found in a PI/Hoechst staining assay
(data not shown).
Alteration of Stat1 gene expression after HDAC
inhibition
We employed microarray and Western blot analyses to
define alterations in gene expression patterns after incu-
bation with HDACi. These assays revealed a time- and
dose-dependent increase in Stat1 expression at the
mRNA and protein level in SK-37 (Fig. 2A) and several
other HDACi-sensitive cell lines. Treatment of SK-37
cells with HDACi and cycloheximide showed that the
HDACi-induced increase in Stat1 expression was depen-
dent on de novo protein synthesis (data not shown).
Hence, an increase in Stat1 stability due to reduced
HDAC activity cannot account for higher Stat1 expres-
sion levels. Intriguingly, HDACi-resistant cell lines,
such as NW-450 and NW-1539, did not undergo HDACi-
induced caspase 3 cleavage and apoptosis (Fig. 1A,B) and
expressed very low levels of Stat1, which were not in-
duced by HDACi (Fig. 2B,C). Since no significant differ-
ence in HDACi-induced histone hyperacetylation was
detected between NW-1539 and SK-37 cells, HDACi
were equally effective in blocking HDAC activity in
both cell lines (Fig. 2C). This result indicates that not
only inhibition of HDACs but also the presence of Stat1
is crucial for HDACi-mediated apoptosis in melanoma
cells. Consistent with previous reports (Wong et al.
2002), we detected a strong increase in Stat1 expression
in the Stat1-positive SK-37 cell line, but not in NW-1539
cells treated with interferon ?. Moreover, cotreatment
with HDACi further increased Stat1 expression in SK-37
cells (data not shown). This correlated with enhanced
induction of apoptosis as assessed by MTT and FACS
analysis (Fig. 2D).
These results prompted us to investigate whether
Stat1? is required for HDACi-induced apoptosis in NW-
1539 cells. We transduced these cells with a lentiviral
vector expressing Stat1?. Indeed, sensitivity toward
HDACi was conferred to Stat1?-transduced NW-1539
cells but not to cells which received only the vector en-
coding GFP (Fig. 3A,B; data not shown). Hence, Stat1
expression levels appear to determine the response of
this cell line to HDACi. Furthermore, introduction of
Stat1? renders these cells susceptible to enhanced apo-
ptosis induction by VPA and interferon ? (Fig. 3B). How-
ever, interferon ? alone did not induce apoptosis and
consistently did not induce Stat1 in this cell line. Similar
results were obtained with the Stat1-negative cell line
U3A (Müller et al. 1993) reconstituted with Stat1?, al-
beit with less pronounced apoptosis induction (data not
shown).
Our results clearly show that HDACi induce Stat1 in
a cell type-specific manner. However, microarray analy-
ses gave no evidence for increased expression of Stat1
target genes as a result of HDAC inhibition in SK-37
cells. Therefore, the activity of Stat1 in HDACi-induced
Figure 2.
tion. (A) The time- and dose-dependent increase of Stat1 expres-
sion was investigated by Western blot. SK-37 cells were exposed
to 1.5 mM VPA or 100 nM TSA for the indicated periods of time.
Alternatively, cells were treated for 24 h with different concen-
trations of VPA (0.1–1.5 mM) or TSA (10–300 nM) as indicated
or left untreated (0). (B) Expression of Stat1? in SK-37 and NW-
450 melanoma cells treated with 1.5 mM VPA (V) or 100 nM
TSA (T) for 24 h or left untreated (C) was analyzed by Western
blot. (C) Expression of Stat1 and accumulation of hyperacety-
lated histone H4 (AcH4) in SK-37 and NW-1539 cells were ana-
lyzed after 24 h by Western blot. Cells were treated with VPA
(V, 1.5 mM) or TSA (T, 30 nM) or left untreated (C). (D) Sensi-
tivity of melanoma cell lines to VPA (V, 1.5 mM) and interferon
? (?, 103U/mL) was determined by MTT assay. Enhanced in-
duction of apoptosis after treatment of SK-37 cells with VPA
and interferon ? (?/V) was detected by PI FACS analysis.
Correlation of Stat1 expression and apoptosis induc-
Acetylation of Stat1
GENES & DEVELOPMENT475

Page 4

apoptosis is likely to involve nongenomic effects, such
as cross-talk with other signaling pathways.
HDACi modulate NF-?B activity
To identify such signaling pathways, gene expression
analyses can provide valuable information. Since a criti-
cal role of NF-?B for HDACi-induced apoptosis has been
described, we analyzed the expression of NF-?B-depen-
dent genes in the HDACi-sensitive cell line SK-37. Our
data indicate HDACi-dependent repression of NF-?B tar-
get genes such as Bcl-XL, survivin, and Stat5 (Fig. 4A).
These results confirm several reports describing effects
of HDACi on these genes (Eickhoff et al. 2000; Krämer et
al. 2001; Hinz et al. 2002; De Schepper et al. 2003). On
the other hand, expression of NF-?B-regulated genes re-
mained unaltered in HDACi-resistant NW-1539 cells,
which expressed hardly any Stat1 (Fig. 4A). Treatment
with the cytokine interferon ? similarly led to repression
of these genes in SK37 but not in NW-1539 cells (Fig.
4A).
Data shown in Figures 2C and 4A indicate that Stat1?
expression inversely correlates with the activity of NF-
?B after HDACi treatment or interferon ? stimulation.
Therefore, we employed a binding assay with a biotinyl-
ated NF-?B consensus site oligonucleotide to investigate
whether DNA binding of NF-?B might be affected in
cells expressing different amounts of Stat1. Decreased
affinity of NF-?B for a cognate DNA sequence was ob-
served after treatment with HDACi or interferon ? only
in cells expressing Stat1 (Fig. 4B). EMSAs confirmed
functional impairment of NF-?B p65/p50 binding to its
DNA sequence in extracts from HDACi-treated Stat1-
positive SK-37 cells (Fig. 4C, left). EMSAs performed
with the same lysates and an oligonucleotide specific for
Stat3 showed no decrease in DNA binding, which rules
out nonspecific effects of HDACi treatment on the DNA
binding of transcription factors (data not shown). Re-
duced affinity of NF-?B for DNA after HDAC inhibition
was also observed with Stat1?-transduced NW-1539
cells, but not in parental or vector-transduced Stat1-
negative NW-1539 cells (Fig. 4C, cf. lanes 5,9,13 and
6,10,14). Hence, DNA binding of NF-?B after treatment
with HDACi is only reduced in the presence of Stat1?,
suggesting a link of both signaling pathways.
We challenged the proposed role of Stat1 in regulating
NF-?B function by transduction of SK-37 cells with len-
tiviral vectors encoding small interfering RNA (siRNA)
directed against Stat1 or a nonrelevant sequence. We ob-
served that basal and induced Stat1 levels were reduced
only in cells that received the siStat1 vector. In control-
transduced SK-37 cells, HDACi or interferon ? still de-
creased the expression of NF-?B target genes as well as
NF-?B DNA binding. However, NF-?B functions were no
longer affected under identical conditions in SK-37 cells
subjected to a constitutive siRNA-mediated knock-
down of Stat1 levels (Fig. 4D,E). In a rescue experiment
we transfected a mutated siRNA-resistant Stat1 vector
into the Stat1 knock-down SK-37 cells. Consequently, a
strong repression of NF-?B functions by HDACi and in-
terferon ? could again be induced. Moreover, high levels
of Stat1 expression caused a reduction of NF-?B activity
independent of these substances. These data demon-
strate that Stat1 is a crucial regulator of NF-?B in vivo
(Fig. 4F,G).
The localization of nuclear p65 is modulated by Stat1
and HDACi
We gained further insights into the mechanism underly-
ing the regulation of NF-?B in SK-37 cells by analyzing
its subcellular distribution. In situ immunofluorescence
analysis showed a shift of NF-?B p65 from the nucleus to
the cytosol and an increased colocalization with Stat1?
in the cytosol when HDACs were inhibited or cells were
treated with interferon ? (Fig. 5A; data not shown).
Treatment of cells with the nuclear export inhibitor
LMB prevented basal and HDACi-induced export of NF-
?B p65 and caused nuclear accumulation of Stat1? and
p65. Again, colocalization of these proteins was en-
hanced (Fig. 5A).
Figure 3.
and interferon ?. (A) Western blot analysis was employed to
detect Stat1 expression and induction of apoptosis in NW-1539
cells transduced with SIEW (vector) or S-Stat1?-IEW (Stat1) and
treated with VPA (1.5 mM) for 48 h. Asterisks denote activated
forms of full-length caspase 3 or caspase 8. (B) DNA fragmen-
tation was analyzed by PI FACS analysis after treatment with
VPA (1.5 mM) or VPA and interferon ? (103U/mL) for 60 h. (Ctl)
Untreated cells.
Stat1 sensitizes resistant melanoma cells to HDACi
Krämer et al.
476 GENES & DEVELOPMENT

Page 5

The analysis of cytosolic and nuclear fractions of SK-
37 cells by Western blot indicates that Stat1 expression
increased both in the cytosol and in the nucleus (Fig. 5B).
In these cells, as in many other cancer cell lines, low
constitutive amounts of NF-?B p65 reside in the nucleus
and help to attenuate proapoptotic signaling. For p65 a
clear reduction in the nuclear compartment was ob-
served after treatment with HDACi as well as interferon
?, which confirms our microscopy results. These data
suggest that removal of p65 from the nucleus and inter-
action with Stat1 might be key steps in the HDACi-
induced repression of NF-?B target genes and apoptosis
induction. Another possible explanation, changes in NF-
?B p65 expression, can be ruled out since p65 levels were
not altered significantly (Fig. 5B). Western blot analysis
of 2fTGH cells and their derived Stat1-negative cell line
U3A confirmed the dependence on Stat1 for nuclear ex-
port of p65 upon HDAC inhibition (Fig. 5C). Hence, it
appears plausible that HDACi and interferon ? inhibit
nuclear localization of p65 only in cells expressing Stat1.
HDACi induce the interaction of Stat1? and NF-?B
Having established a role of Stat1? in NF-?B signaling,
we investigated whether these transcription factors
could interact physically. First, we tested if this interac-
tion is mediated by TRADD, which was shown to asso-
ciate with Stat1? under certain conditions (Wang et al.
2000). However, in several immunoprecipitation (IP) ex-
periments, we could not detect an HDACi-dependent in-
teraction of Stat1? with TRADD in SK-37 cells (data not
shown). On the other hand, precipitation of Stat1? or
NF-?B p65 revealed an association of these proteins upon
HDAC inhibition or interferon stimulation (Fig. 5D). Co-
immunoprecipitation(co-IP)
with cytosolic and nuclear fractions confirmed our mi-
croscopy data showing that such complexes are located
in the cytosol. Moreover, Stat1–NF-?B complex forma-
tion is likely to be DNA independent, since addition of
ethidium bromide did not alter its stability (data not
shown). Next, we analyzed Stat1 complexes by Superose
6 column fractionation of SK-37 cell extracts and found
that in high molecular weight fractions the amount of
Stat1 increased together with NF-?B p65 after VPA treat-
ment (Fig. 5E). Notably, this complex accumulated in a
time-dependent manner that paralleled apoptosis induc-
tion after HDAC inhibition. IP of Stat1? from these frac-
tions followed by Western blotting against p65 showed
that a weak basal interaction of these proteins increased
upon HDAC inhibition (Fig. 5E).
We also analyzed the presence of HDACs in the Stat1
complex before and after HDAC inhibition or treatment
with interferon ? by IP and Western blot. The addition of
HDACi to SK-37 cells led to reduced association of Stat1
experimentsperformed
Figure 4.
Expression of NF-?B target genes after HDAC inhi-
bition and treatment with interferon ? was investi-
gated by Western blot analysis of SK-37 and NW-
1539 cell lysates. Cells were incubated with 1.5 mM
VPA (V), 30 nM TSA (T), or 103U interferon ? (IFN),
or left untreated (C) for 48 h. (B) ABCD-assay with a
biotinylated NF-?B consensus oligo was used to de-
tect NF-?B–DNA binding under conditions de-
scribed in A. (N) Nonrelevant biotinylated oligos.
(C) NF-?B DNA binding was analyzed by EMSA of
lysates from SK-37, NW-1539, and transduced NW-
1539 cells (vector or Stat1) that were either un-
treated or treated with VPA (1.5 mM) for 48 h. Iden-
tity of the NF-?B–DNA complex was verified by p65
and p50 antibody supershifts (SS-AB). (D) NF-?B tar-
get gene expression was investigated in lysates of
SK-37 cells transduced with an siRNA vector encod-
ing scrambled RNA or an si-sequence against Stat1.
Experimental conditions are as described in A. (E)
NF-?B–DNA binding was investigated in lysates of
SK-37 cells by ABCD-assay. Cells and conditions are
as described in A. (F) NF-?B target gene expression
was investigated in lysates of SK-37 cells transduced
as stated in D and transfected with empty vector or
pc3 HA-Stat1 containing mutations conferring resis-
tance against the siRNA. Experimental conditions
are as described in A. (G) NF-?B–DNA binding was
investigated in lysates of SK-37 cells by ABCD-as-
say. Cells and conditions are as described in E.
Stat1 interferes with NF-?B function. (A)
Acetylation of Stat1
GENES & DEVELOPMENT477