A cytoplasmic negative regulator isoform of ATF7 impairs ATF7 and ATF2 phosphorylation and transcriptional activity.

Abstract

Alternative splicing and post-translational modifications are processes that give rise to the complexity of the proteome. The nuclear ATF7 and ATF2 (activating transcription factor) are structurally homologous leucine zipper transcription factors encoded by distinct genes. Stress and growth factors activate ATF2 and ATF7 mainly via sequential phosphorylation of two conserved threonine residues in their activation domain. Distinct protein kinases, among which mitogen-activated protein kinases (MAPK), phosphorylate ATF2 and ATF7 first on Thr71/Thr53 and next on Thr69/Thr51 residues respectively, resulting in transcriptional activation. Here, we identify and characterize a cytoplasmic alternatively spliced isoform of ATF7. This variant, named ATF7-4, inhibits both ATF2 and ATF7 transcriptional activities by impairing the first phosphorylation event on Thr71/Thr53 residues. ATF7-4 indeed sequesters the Thr53-phosphorylating kinase in the cytoplasm. Upon stimulus-induced phosphorylation, ATF7-4 is poly-ubiquitinated and degraded, enabling the release of the kinase and ATF7/ATF2 activation. Our data therefore conclusively establish that ATF7-4 is an important cytoplasmic negative regulator of ATF7 and ATF2 transcription factors.

Full text

Page 1

A Cytoplasmic Negative Regulator Isoform of ATF7
Impairs ATF7 and ATF2 Phosphorylation and
Transcriptional Activity
Jessica Diring1, Barbara Camuzeaux1, Mariel Donzeau1, Marc Vigneron1, Manuel Rosa-Calatrava2,
Claude Kedinger1, Bruno Chatton1*
1Universite ´ de Strasbourg, UMR7242 Biotechnologie et Signalisation Cellulaire, Ecole Supe ´rieure de Biotechnologie de Strasbourg, BP10413, Illkirch, France, 2Laboratoire
de Virologie et Pathologie Humaine VirPath, Universite ´ Claude Bernard Lyon 1, Hospices Civils de Lyon, Lyon, France
Abstract
Alternative splicing and post-translational modifications are processes that give rise to the complexity of the proteome. The
nuclear ATF7 and ATF2 (activating transcription factor) are structurally homologous leucine zipper transcription factors
encoded by distinct genes. Stress and growth factors activate ATF2 and ATF7 mainly via sequential phosphorylation of two
conserved threonine residues in their activation domain. Distinct protein kinases, among which mitogen-activated protein
kinases (MAPK), phosphorylate ATF2 and ATF7 first on Thr71/Thr53 and next on Thr69/Thr51 residues respectively, resulting
in transcriptional activation. Here, we identify and characterize a cytoplasmic alternatively spliced isoform of ATF7. This
variant, named ATF7-4, inhibits both ATF2 and ATF7 transcriptional activities by impairing the first phosphorylation event
on Thr71/Thr53 residues. ATF7-4 indeed sequesters the Thr53-phosphorylating kinase in the cytoplasm. Upon stimulus-
induced phosphorylation, ATF7-4 is poly-ubiquitinated and degraded, enabling the release of the kinase and ATF7/ATF2
activation. Our data therefore conclusively establish that ATF7-4 is an important cytoplasmic negative regulator of ATF7 and
ATF2 transcription factors.
Citation: Diring J, Camuzeaux B, Donzeau M, Vigneron M, Rosa-Calatrava M, et al. (2011) A Cytoplasmic Negative Regulator Isoform of ATF7 Impairs ATF7 and
ATF2 Phosphorylation and Transcriptional Activity. PLoS ONE 6(8): e23351. doi:10.1371/journal.pone.0023351
Editor: Pierre-Antoine Defossez, Universite ´ Paris-Diderot, France
Received May 23, 2011; Accepted July 13, 2011; Published August 16, 2011
Copyright: ? 2011 Diring et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study was supported by Universite ´ de Strasbourg, the Centre National de la Recherche Scientifique, the Ministe `re de l’Enseignement Supe ´rieur et
de la Recherche, the Ligue contre le Cancer - Comite ´s Haut-Rhin et Bas-Rhin, the Fondation pour la Recherche Me ´dicale, Alsace contre le Cancer, the Ecos-Sud
(action A06B02), and the Institut National du Cancer (PLBIO-127). The funders had no role in study design, data collection and analysis, decision to publish, or
preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: bchatton@unistra.fr
Introduction
The characterization of cellular pathways leading to all the post-
translational modifications of a protein is essential to understand the
molecular mechanisms regulating its functions. Crosstalks between
different types of such modifications are an emerging theme in
eukaryotic biology. Thus, examples of multiple connections between
phosphorylation, sumoylation and ubiquitination have been described
(for a review see [1]). Within the AP-1 transcription factors family, the
ATF2, ATF7 and CREB5 compose a subfamily based on sequence
conservation [2,3,4,5]. The transcriptional activation and DNA-
binding domainsofATF2 andATF7factors are highlyconserved and
their specificity is mainly governed by post-translational modifications
and interactions with specific cofactors [6,7,8,9]. ATF2 is a protein
that displays diverse, tissue-dependent functions [10,11]. For instance,
ATF2 has been implicated in malignant and non-malignant skin
tumor development [12,13]. ATF2 also elicitsa suppressor function in
mammary tumors [14], and mediates lipopolysaccharide-induced
transcription in macrophage cells [15].
ATF7 shares a considerable sequence homology with ATF2,
especially within the C-terminal DNA-binding/dimerization
domain and the N-terminal activation domain. This latter region
includes a critical zinc-binding element and two conserved
threonine residues (Thr51 and Thr53 corresponding to the
Thr69 and Thr71 homologues in ATF2). Different mitogen-
activated protein kinases (MAPK), particularly members of JNK
and p38 families, specifically phosphorylate these conserved
threonine residues of ATF2 and ATF7 leading to transcriptional
activation [16,17,18,19,20,21,22,23,24,25]. These phosphoryla-
tion events are essential for ATF7/ATF2 proteins function in vivo,
since Atf2A/Amice, carrying mutations in these critical phosphor-
ylation sites, have a strong phenotype, identical to that seen upon
deletion of the DNA-binding domain [26,27]. Combining this
mutant with a knockout of ATF7 (Atf72/2) results in embryonic
lethality with severe abnormalities in the developing liver and
heart [26]. Moreover, Atf72/2mice exhibit abnormal behaviors
and increased serotonin 5-HT receptor 5B (Htr5b) mRNA levels
in the dorsal raphe nuclei [28].
Alternative splicing of pre-mRNAs accounts for a large
proportion of proteomic complexity [29,30]. New splicing events
occurring within protein coding regions potentially alter the
biological function of the gene by expressing polypeptides with
novel functions [31,32]. Until now, the ATF7 transcription factor
subfamily was known to include at least two major proteins (ATF7-
1, ATF7-2) translated from alternatively spliced messengers
transcribed from a unique gene [33]. In the present study, we
report the identification and functional characterization of a short
alternatively spliced variant of ATF7 that is highly conserved
PLoS ONE | www.plosone.org1 August 2011 | Volume 6 | Issue 8 | e23351

Page 2

amongst mammalianspecies, exceptmouseandrat.This messenger
RNA encodes a functional C-terminally truncated protein, ATF7-4,
whose intracellular location is restricted to the cytoplasm and which
downregulatesthetranscriptionalactivityofboth nuclearATF7 and
ATF2 transcription factors. We show that, under resting conditions,
ATF7-4 interacts in the cytoplasm with a kinase. Our data are
compatible with the idea that this kinase is responsible for the
priming phosphorylation event of the ATF7/ATF2 factors. In
contrast, a stimulus-induced phosphorylation of ATF7-4 leads to its
poly-ubiquitination and degradation by a proteasome 26S-
dependent pathway, enabling ATF7/ATF2 activation.
Results
Identification of a new ATF7 isoform
To look for new ATF7 isoforms that could participate in the
ATF7-dependent transcriptional regulation, we performed an-
chored RT-PCR on HeLa mRNA using oligo-(dT) and ATF7-
specific primers overlapping the translation initiation codon. We
obtained a short amplification product (809 bp) that encodes an
ATF7 alternatively spliced isoform of 117 residues designated
ATF7-4 (following the ATF7 nomenclature). However, no such
short product was amplified in mouse 3T3 fibroblasts under the
same conditions (data not shown). To understand the existence of
this new isoform, a comparative analysis of human and mouse
ATF7 gene sequences was performed (Figure 1A). First, we
observed that the human exon 4 is followed by a potential exon,
referred to as exon 4a, that contains an open-reading frame
encoding 29 amino acids. Moreover, this newly identified exon
that is not found in the mouse gene displays a polyadenylation
sequence (AATAAA) at its 39 border. The inclusion of exon 4a is
enabled when the splice donor site (DS) of exon 4 is not used. As a
result, a short transcript encoding ATF7-4 is generated. Interest-
ingly, examination of the nucleotide sequence of the ATF7 gene
also revealed in exon 5 an alternative splice donor site (DS1),
absent in the mouse gene, that can be used in place of the donor
site of the exon 5a (DS2) (Figure 1A). The use of DS1 gives rise to
the human ATF7-1 isoform. By contrast, the ATF7-2 variant
encompassing exons 5 and 5a is expressed in both human and
mouse. Thus, three major different transcripts are encoded in
Figure 1. Identification of ATF7-4 as a novel alternatively spliced ATF7 isoform. (A) The structure of the ATF7 gene, a member of the ATF2/
ATF7/CREB5 family, is shown schematically: the twelve exons are indicated by numbered boxes, and an alternative splicing region encompassing
exons 4 to 6 is compared between human and mouse species. The sequences of the donor (DS) and acceptor (AS) sites for splicing are indicated,
capital and small letters corresponding to the exonic and the intronic parts respectively. 4a and 5a are alternatively included exons adjacent to exons
4 and 5 respectively. The start codon, stop codons and polyadenylation sites are shown. (B) Schematic representation of the exonic structure of three
ATF7 alternatively spliced isoforms. The regions encoding the zinc finger (ZnF)/activation domain (AD) or the nuclear localization signal (NLS)/basic
region-leucine zipper (b-Zip) are indicated. ‘‘+ +’’ indicates that the proteins are encoded, either in human or in mouse species. (C) Amino acid sequence
alignment of ATF7-4 exon 4 and specific exon 4a of a series of mammalian species. A conserved C-terminal hydrophobic domain is highlighted.
doi:10.1371/journal.pone.0023351.g001
ATF7-4 Inhibits ATF7 and ATF2 Activation
PLoS ONE | www.plosone.org 2August 2011 | Volume 6 | Issue 8 | e23351

Page 3

human (ATF7-1, ATF7-2, ATF7-4) whereas only one (ATF7-2) is
encoded in mouse (Figure 1B).
The ATF7-4 protein contains the common ATF7 N-terminal
moiety (88 residues) encompassing the transcriptional activation
domain [34], and a 29-amino acid extension fused in frame.
However, it lacks the entire C-terminal basic region and leucine
zipper domain (residues 89–494) that function in ATF7-1 and
ATF7-2 as a DNA binding domain/nuclear localization signal
(NLS) and as a dimerization domain respectively [33] (Figure 1B).
It is likely therefore that ATF7-4 exhibits functions that are distinct
from those of the ATF7 full-length (ATF7-FL) isoforms. The
alignment of peptide sequences derived from genomic sequences
available in the databases reveals that the ATF7-4 isoform is
potentially encoded in many mammalian genomes but not in
mouse or rat. Moreover, it exhibits a highly conserved hydropho-
bic domain (HD) in its C-terminal moiety (Figure 1C).
ATF7-4 is ubiquitously but differentially transcribed
To investigate the expression pattern of the ATF7-4 isoform,
cDNAs panels from human tissues and different cell lines were
analyzed by qRT-PCR using primers that are specific of the
different species of ATF7 cDNAs (Table S1). We observed that
ATF7-4 and ATF7-FL (i.e. ATF7-1 and ATF7-2) transcripts are
detectable in all the human tissues and cell lines tested, reflecting a
ubiquitous expression (Figure 2A and B, left panels, and Figure
S2). However, the ATF7-4 transcript is predominantly detected in
brain, heart, kidney, muscle, testis and also in HeLa, A549 and
H1299 cell lines (Figure 2A and B, left panels). As expected, no
ATF7-4 expression was detected in mouse embryonic fibroblasts
(MEF, data not shown). We then assessed the relative ratio of
ATF7-4 mRNA levels compared to those of ATF7-FL in the same
tissues and cell lines. Interestingly, this analysis revealed a rather
strong disparity of the relative accumulation levels of the two
mRNA species, ATF7-4 mRNAs being two to seven times more
abundant in some organs, i.e. kidney, muscle, thyroid, and nearly
absent in pancreas (Figure 2A and B, right panels).
ATF7-4 protein level was next analyzed by western-blot in
HeLa and 3T3 cell extracts (Figure 2C). The monoclonal antibody
used in these experiments was directed against a peptide spanning
residues 96–108, a region unique to the ATF7-4 protein. As
expected, ATF7-4 protein expression was detected in HeLa but
not in mouse 3T3 cell extracts (Figure 2C). Moreover, a specific
siRNA efficiently knocked down ATF7-4. Together, these results
indicate that human ATF7-4 is ubiquitously but differentially
expressed, potentially suggesting a tissue-specific role for this
variant. The high ATF7-4 mRNA and protein levels detected in
HeLa cells led us to study its cellular function in those cells.
ATF7-4 is localized in the cytoplasm
Since ATF7-4 essentially comprises the activation domain of the
ATF7 proteins and lacks their C-terminal dimerization/nuclear
localization signal (NLS) domain, we examined its subcellular
localization compared to those of the ATF7-FL proteins.
Immunofluorescent staining experiments with specific ATF7
antibodies revealed that overexpressed ATF7-4 protein mainly
localizes in the cytoplasm, whereas ATF7-FL proteins are nuclear
in HeLa cells (Figure 3A). In an attempt to map the peptidic
elements mediating ATF7-4 cytoplasmic localization, the hydro-
phobic domain (HD) that is specific to ATF7-4 and absent in
ATF7-FL was fused, either in its wild-type or mutant version, at
the C-terminal end of the eGFP (Figure 3B). In the mutant
derivative, the hydrophobic domain was disrupted by replacing
four leucine and isoleucine residues with alanine residues
(Figure 3B). These eGFP-fusion proteins were then expressed in
HeLa cells, and their localization was assessed by immunofluo-
rescence (Figure 3C). A quantification of the nuclear fluorescence
intensity over the total cell intensity was performed for each
construct (Figure 3D). Control experiments indicated that the
eGFP protein was expressed both in nucleus and cytoplasm
(Figure 3C and D, row 1), whereas it was clearly directed to the
cytoplasm when fused to the ATF7-4 specific domain (Figure 3C
and D, row 2). Moreover, the nuclear localization was rescued by
the mutations in the hydrophobic domain (Figure 3C and D, row
3). These results clearly demonstrate that the ATF7-4 transcription
factor isoform is cytoplasmic, and that this localization is mediated
by its C-terminal hydrophobic domain.
ATF7-4 is phosphorylated and subsequently degraded by
a proteasome-dependent pathway
We have previously demonstrated that both basal and p38b2
MAP kinase-stimulated ATF7-FL transcriptional activities depend
mainly on the phosphorylation of residues Thr51, Thr53 and
Thr112 within the activation domain of the protein, although the
p38b2-mediated phosphorylation is restricted to Thr51 [23]. Since
ATF7-FL proteins and the cytoplasmic variant ATF7-4 share the
same N-terminal part of the activation domain, we investigated the
contribution of the two Thr51 and Thr53 residues in both basal
and p38b2-mediated ATF7-4 phosphorylation. Site-directed
mutagenesis was performed to modify these two residues into
alanine residues, generating the T51A and T53A mutants
(Figure 4A). The wild-type or mutated versions of ATF7-4 were
co-expressed or not with MKK6-activated p38b2in HeLa cells
and analyzed by western-blotting using specific antibodies. The
monoclonal antibody directed against the C-terminal part of
ATF7-4specifically recognizes
(Figure 4B, upper panel, and data not shown). The phospho-
specific antibodies targeting ATF7 when it is phosphorylated on
Thr51 (pThr51) and Thr53 (pThr53) have been previously
described [23] (Figure 4B, lower panels). In the absence of
activated p38b2, the phosphorylation of Thr53 residue could be
revealed using the pThr53 antibody (Figure 4B, lane 1). However,
no signal was detected with the pThr51 antibody, indicating that,
like the ATF7-FL protein, ATF7-4 is not phosphorylated on
residue Thr51 under these conditions. By contrast, co-expression
of activated p38b2led to Thr51 phosphorylation (Figure 4B, lane
2). This double phosphorylated form (Thr51/Thr53) was more
efficiently recognized by the pThr53 antibody, as previously
described [23], resulting in an increased signal (Figure 4B, lane 2).
The mutation of each of the two threonine residues affected the
phosphorylation pattern of the ATF7-4 protein (Figure 4B, lanes
3–6). Importantly, the T53A mutation impaired Thr51 phosphor-
ylation (Figure 4B, lane 6), indicating that Thr53 integrity is
required for subsequent Thr51 phosphorylation, as it is the case
for ATF7-FL that exhibits a two-step phosphorylation mechanism
[23]. That the ATF7-4 species detected by the phospho-specific
antibodies correspond to phosphorylated forms is further demon-
strated by their sensitivity to calf intestinal phosphatase (CIP)
treatment and their rescue in presence of EDTA, an inhibitor of
CIP (Figure 4B, lanes 7 and 8). In conclusion, the ATF7-4 protein
is phosphorylated in a two-step manner like ATF7-FL, first on
Thr53 and then on Thr51 by the p38b2MAP kinase.
We next studied the stability of ATF7-4 and ATF7-FL proteins
by using MG-132, a proteasome 26S inhibitor. ATF7-4 (Figure 4C,
lanes 1–4) or ATF7-FL (Figure 4C, lanes 5–8) proteins were co-
expressed or not with activated p38b2in HeLa cells in the absence
or presence of MG-132. We then compared the respective
amounts of the non-phosphorylated and phosphorylated ATF7-4
and ATF7-FL proteins in these conditions. After MG-132
non-phosphorylated ATF7-4
ATF7-4 Inhibits ATF7 and ATF2 Activation
PLoS ONE | www.plosone.org3August 2011 | Volume 6 | Issue 8 | e23351

Page 4

treatment, we observed a nearly ten fold enrichment of both basal
phosphorylated (Figure 4C, lane 2) and p38b2-phosphorylated
(Figure 4C, lane 4) ATF7-4 proteins in cell extracts, clearly
indicating that phosphorylated ATF7-4 is stabilized after protea-
some inhibition. In contrast, no difference could be observed for
non-phosphorylated ATF7-4 (Figure 4C, lanes 1–4, upper panel)
or ATF7-FL (Figure 4C, lanes 5–8), whether phosphorylated or
not, under the same conditions. Thus, it appears that the ATF7-4
isoform is specifically prone to proteasome-dependent degradation
once phosphorylated, in contrast to ATF7-FL.
To further confirm the implication of the proteasome in the
phospho-ATF7-4 degradation, we co-expressed ATF7-4 and
His-tagged ubiquitin in HeLa cells treated with MG-132. His-
tagged proteins, i.e. ubiquitin and ubiquitin-conjugated proteins,
were purified on nickel beads and analyzed by immunoblotting
with specific antibodies. Rabbit antiserum recognizing phosphor-
ylated ATF7-4 (pATF7-4) enabled us to detect a specific pattern of
poly-ubiquitinated phospho-ATF7-4 proteins (Figure 5D, lane 2).
These results demonstrate that phosphorylated ATF7-4 are
substrates for poly-ubiquitination.
Taken together, these results strongly suggest that ATF7-4 is
specifically degraded via the ubiquitin-proteasome 26S pathway
when it is phosphorylated, whereas ATF7-FL is not.
Figure 2. Pattern of expression of ATF7-4 transcript. (A and B) Real-time RT-PCR analysis of ATF7-4 gene expression profile (left panels) (A) in a
24 human tissue-cDNA array and (B) in a selection of human transformed cell lines. Right panels show the ATF7-4 expression relative to that of ATF7-
FL. PBL stands for peripheral blood leucocytes. Data are the average of at least five independent experiments (standard deviations are shown). Data
were normalized to b-actin gene expression. (C) Endogenous ATF7-4 protein was analyzed by western-blotting (WB) with a specific anti-ATF7-4
monoclonal antibody in human HeLa cells and mouse 3T3 fibroblasts. HeLa cells were transfected with either the px-ATF7-4 expression vector or a
control, or specific siRNA targeting ATF7-4. b-actin was analyzed in parallel as a loading control.
doi:10.1371/journal.pone.0023351.g002
ATF7-4 Inhibits ATF7 and ATF2 Activation
PLoS ONE | www.plosone.org4August 2011 | Volume 6 | Issue 8 | e23351

Page 5

ATF7-4 downregulates the transcriptional activity of
ATF7-FL and ATF2
We looked for a functional role of ATF7-4 isoform. Since
ATF7-4 and ATF7-FL proteins share the same activation domain,
but are located in two distinct cellular compartments, we tested
whether ATF7-4 could have an effect on ATF7-FL transcriptional
activity or act as a negative regulator form of ATF7-FL. To answer
this question, we assessed the transcriptional activity of ATF7-FL
under various conditions. In order to circumvent a potential
interference of the endogenous ATF proteins, we used an assay
based on a luciferase reporter controlled by Gal4 fusion proteins
[35]. ATF7 and ATF2 N-terminal sequences were fused to the
Gal4 DNA-binding domain (Figure 5A, a). The resulting fusion
proteins were co-expressed in HeLa cells with increasing amounts
of ATF7-4 wild-type (WT) or mutant versions (Figure 5A, b), and
their transcriptional properties were monitored by measuring the
luciferase activity in the cellular extracts. The ATF7-4 T51A
T53A and ATF7-4 C9A T51A T53A derivatives are non-
phosphorylable proteins, while the latter exhibits an additional
mutation in the zinc finger motif that prevents the binding of
different ATF7 partners, among which its TAF12 coactivator [8].
The activity of the Gal4-ATF7(2–82) moiety was clearly inhibited
in the presence of increasing amounts of ATF7-4 WT and to a
lesser extent by ATF7-4 T51A T53A, whereas it was unaffected by
the ATF7-4 C9A T51A T53A protein (Figure 5B).
To determine whether this phenomenon is specific to ATF7 or
has a wider impact, we then analyzed the effect of ATF7-4 on the
transcriptional activity of ATF2 transcription factor, structurally
and functionally related to ATF7. The impact of ATF7-4 on the
transcriptional activities of Gal4-ATF2 and the unrelated Gal4-
Elk1 factor as a control, was assessed by luciferase assays.
Interestingly, ATF7-4 expression resulted in a strong inhibition
0,0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
Figure 3. The hydrophobic domain of ATF7-4 controls its cytoplasmic localization. (A and C) Confocal microscopy images of (A) ATF7-FL,
ATF7-4, and (C) eGFP proteins or eGFP-ATF7-4 fusion versions overexpressed in HeLa cells and stained by immunofluorescence with specific
antibodies (central panels). DNA was stained with Hoechst 33258 (left panels). A merge of signals is shown on the right panels. (B–D) Analysis of the
functional role of the ATF7-4 hydrophobic domain. (B and C) The relocalization of eGFP (lane 1) was assessed for eGFP fusion proteins with ATF7-4
hydrophobic domain (HD), either wild type (wt, lane 2) or mutant version (mt, lane 3). In the latter construct, the four conserved hydrophobic
residues (highlighted) were mutated (asterisks) to alanine. Representative confocal images are shown. (D) The fluorescence intensity in the nucleus
(N) over the total cell intensity (N+C) was quantified using ImageJ. The results presented are the average of at least 22 cells analyses for each
condition. Standard deviations and statistical significance by Student t-test are shown: NS=nonsignificant, ***p,0.001. Scale bar: 10 mm.
doi:10.1371/journal.pone.0023351.g003
ATF7-4 Inhibits ATF7 and ATF2 Activation
PLoS ONE | www.plosone.org5August 2011 | Volume 6 | Issue 8 | e23351