Identification of mitogen-activated protein⁄extracellular
signal-responsive kinase kinase 2 as a novel partner of the
scaffolding protein human homolog of disc-large
Oumou Maı
¨ga
1
, Monique Philippe
1
, Larissa Kotelevets
2
, Eric Chastre
2
, Samira Benadda
3
,
Dominique Pidard
1
, Roger Vranckx
1
and Laurence Walch
1
1 INSERM U698, Universite
´Paris 7, France
2 INSERM U773, Centre de Recherche Biome
´dicale Bichat Beaujon, Paris, France
3 Plateau de Microscopie Confocale ICB-IFR 02, Paris, France
Keywords
human disc-large homolog; human vascular
smooth muscle cells; MAPK ERK kinase 2;
scaffold protein; synapse-associated
protein 97
Correspondence
L. Walch, INSERM U698, Cardiovascular
Haematology, Bio-Engineering and
Remodelling, Bichat-Claude Bernard
Hospital, 46 rue Henri Huchard, F-75877,
Paris, Cedex 18, France
Fax: +33 1 40 25 86 02
Tel: +33 1 40 25 75 22
E-mail: laurence.walch@inserm.fr
(Received 11 January 2011, revised 29 April
2011, accepted 20 May 2011)
doi:10.1111/j.1742-4658.2011.08192.x
Human disc-large homolog (hDlg), also known as synapse-associated
protein 97, is a scaffold protein, a member of the membrane-associated
guanylate kinase family, implicated in neuronal synapses and epithelial–
epithelial cell junctions whose expression and function remains poorly char-
acterized in most tissues, particularly in the vasculature. In human vascular
tissues, hDlg is highly expressed in smooth muscle cells (VSMCs). Using
the yeast two-hybrid system to screen a human aorta cDNA library, we
identified mitogen-activated protein ⁄extracellular signal-responsive kinase
(ERK) kinase (MEK)2, a member of the ERK cascade, as an hDlg binding
partner. Site-directed mutagenesis showed a major involvement of the
PSD-95, disc-large, ZO-1 domain-2 of hDlg and the C-terminal sequence
RTAV of MEK2 in this interaction. Coimmunoprecipitation assays in both
human VSMCs and human embryonic kidney 293 cells, demonstrated that
endogenous hDlg physically interacts with MEK2 but not with MEK1.
Confocal microscopy suggested a colocalization of the two proteins at the
inner layer of the plasma membrane of confluent human embryonic kidney
293 cells, and in a perinuclear area in human VSMCs. Additionally, hDlg
also associates with the endoplasmic reticulum and microtubules in these
latter cells. Taken together, these findings allow us to hypothesize that
hDlg acts as a MEK2-specific scaffold protein for the ERK signaling path-
way, and may improve our understanding of how scaffold proteins, such
as hDlg, differentially tune MEK1 ⁄MEK2 signaling and cell responses.
Structured digital abstract
lhDlg and MEK2 colocalize by fluorescence microscopy (View Interaction 1,2,3)
lhDlg physically interacts with MEK2 by two hybrid (View Interaction 1,2,3)
lhDlg physically interacts with MEK2 by anti bait coimmunoprecipitation (View Interac-
tion 1,2)
lMEK2 physically interacts with hDlg by anti bait coimmunoprecipitation (View Interac-
tion 1,2)
Abbreviations
CHO, Chinese hamster ovary; ERK, extracellular signal-responsive kinase; GK, guanylate kinase; hDlg, human disc-large homolog; HEK-293,
human embryonic kidney 293; hVSMC, human vascular smooth muscle cell; MAGUK, membrane-associated guanylate kinase; MAPK,
mitogen-activated protein kinase; MEK1 ⁄2, MAPK ERK kinase 1 ⁄2; PDZ, PSD-95, disc-large, ZO-1.
FEBS Journal 278 (2011) 2655–2665 ª2011 The Authors Journal compilation ª2011 FEBS 2655
Introduction
The mitogen-activated protein kinases (MAPKs) are a
family of S ⁄T-protein kinases, including p38, c-Jun
N-terminal kinase and extracellular signal-responsive
kinase (ERK)1 ⁄2, which control several biological
processes such as proliferation, differentiation, survival
and apoptosis. The ERK signaling pathway includes
three major components that are activated in cascade
by phosphorylation. Raf phosphorylates two serine
residues in the activation loop of mitogen-activated
protein ⁄ERK kinase (MEK)1 ⁄2. MEK 1 ⁄2 phosphory-
lates ERK1 ⁄2 on both the threonine and tyrosine resi-
dues in the conserved TEY sequence [1] and activated
ERK phosphorylates the serine or threonine residues
on the S ⁄T-P consensus site in more than 100 nuclear,
cytosolic or membrane substrates with diverse func-
tions [2]. The outcomes of ERK activation are as vari-
ous as the ERK substrates, and so an accurate
regulation of the ERK signaling pathway is necessary.
This pathway is under the control of different regula-
tory elements such as phosphatases, docking domains
and scaffold proteins [2–4]. Docking domains are
consensus sequences that MAPK recognize both on
their substrates, as well as on relevant down-regulating
phosphatases and scaffold proteins [4]. The latter can
be divided into two categories [2]. Upstream scaffold
proteins interact with at least one MAPK implicated
in ERK activation to facilitate a functional interaction
and regulate the localization and the duration of the
signal. For example, the MEK partner 1 directs the
ERK cascade to the surface of the late endosomes [5].
Downstream scaffold proteins bind ERK and direct it
to specific substrates. For example, the phosphoprotein
enriched in astrocyte-15 binds ERK1 ⁄2 and ribosomal
protein S6 kinase 2, a direct substrate of ERK, thereby
enhancing the activation of this latter kinase [6].
Human disc-large homolog (hDlg) is a member of
the membrane-associated guanylate kinase (MAGUK)
scaffold protein family [7]. Interaction with MAGUK
permits the formation of multiprotein complexes, sta-
ble subcellular localizations of interacting partners and
the coordination of their activities. MAGUK contain a
number of protein–protein interaction domains, such
as PSD-95, disc-large, ZO-1 (PDZ), Src-homology 3
and guanylate kinase (GK) domains. In particular,
PDZ domains contain a specific GLGF sequence that
constitutes a hydrophobic cavity where the X-S ⁄T-X-
V⁄L C-terminal motif of their target proteins binds [8].
hDlg expression has been established in a variety of
cells, including neurons, astrocytes, epithelial cells and
T lymphocytes, where hDlg interacts with cytoskeleton
proteins, ion channels, receptors or signaling proteins,
such as kinases. The association of hDlg with kinases
allows the orchestration of cell-specific signaling path-
ways. For example, hDlg ⁄p38 association coordinates
T cell receptor signaling in T lymphocytes, whereas
hDlg recruits phosphatidylinositol 3-kinase to E-cadh-
erin complexes, allowing integrity of the adherent junc-
tion in epithelial cells [9,10]. It should be noted that
there has been no demonstration to date showing that
MAGUK are implicated in the ERK cascade.
Little is known about the role of hDlg in the cardio-
vascular system. Previous studies have shown that
hDlg is expressed in the myocardium where it can form
complexes with K
+
channels such as the inwardly-rec-
tifying K
+
channel 2.2 or the voltage-gated K
+
chan-
nel, allowing functional channel clustering and an
enhancement of the K
+
current [11–13]. However, the
putative expression and functions of hDlg remain to be
established in vascular tissues. To gain insight into
hDlg expression and specific functions in human vascu-
lar tissues, we examined hDlg expression in human
arteries and, more particularly, in human vascular
smooth muscle cells (hVSMCs), and searched for PDZ
domain-dependent binding partners. A screening of a
human aorta cDNA library by the yeast two-hybrid
assay allowed us to identify MEK2 as a new potential
binding partner for hDlg. This interaction was then
validated by biochemical procedures, including coim-
munoprecipitation and confocal immunomicroscopy
colabeling using cultured hVSMCs, as well as Chinese
hamster ovary (CHO) and human embryonic kidney
293 (HEK-293) cells as models.
Results
hDlg protein is present in hVSMCs
As shown in Fig. 1A, immunohistochemical labeling of
hDlg carried out on sections of nonpathological
human mammary arteries revealed that, among the
arterial tissue layers, the media specifically exhibited a
strong staining. Because VSMCs are the only cell type
found in the healthy arterial media, hDlg expression
and subcellular localization were then investigated in
cultured primary hVSMCs. Immunoblot analysis of
total or subcellular protein extracts prepared from con-
fluent hVSMCs identified the presence of two molecu-
lar immunoreactive species both in the total extract
and in the membrane fraction (Fig. 1B), whereas they
were absent in the cytosolic fraction. Taken together,
these data suggest that hDlg is associated with
membrane components. By immunofluorescent labeling
hDlg in ERK cascade O. Maı
¨ga et al.
2656 FEBS Journal 278 (2011) 2655–2665 ª2011 The Authors Journal compilation ª2011 FEBS
coupled with confocal microscopy, hDlg was observed
to be widely distributed within the cytoplasm of
hVSMCs (Fig. 1C–E). Costaining with various orga-
nelle markers showed that hDlg partially colocalized
with endoplasmic reticulum-associated calreticulin
(Fig. 1C), with Golgi-associated GM130 (Fig. 1D) and
with tubulin at the cell periphery, as well as in the
cytoplasm (Fig. 1E), and locally with cortical F-actin
(Fig. 1F). Taken together, these data suggest that hDlg
is mainly associated with internal membrane structures
and with the cytoskeleton in hVSMCs.
Two hDlg isoforms are expressed in human
arteries
hDlg mRNAs are known to contain three regions that
encompass alternatively spliced exons (Fig. 2A), lead-
ing to several hDlg isoforms [14,15]. To further charac-
terize hDlg isoforms expressed in hVSMCs, primer
pairs were chosen within the exons that surround the
region of alternative splicing (Fig. 2A) and RT-PCR
experiments were carried out on human de-endothelial-
ized pulmonary arterial RNA extracts (Fig. 2B–D),
latoT
lo
s
otyC
senarb
m
e
M
Ct
200 µm
hDlg
BA
C
D
E
F
hDlg
N-cadherin
RSK
Fig. 1. Detection of hDlg in hVSMCs. (A)
Serial frozen sections of human mammary
artery were stained with monoclonal hDlg
antibody or an irrelevant mouse IgG
1
(Ct).
(B) Total hVSMC lysate, or a lysate fraction-
ated into membrane-associated and cyto-
solic proteins, was submitted to western
blot detection of hDlg and the fraction
markers N-cadherin and ribosomal S6 kinase
(RSK). (C–F) Cultured hVSMCs were stained
for hDlg (green signal) and, as a red signal,
(C) calreticulin, an endoplasmic reticulum
marker, (D) GM130, a Golgi marker, (E)
tubulin or (F) F-actin. Cells were analyzed by
confocal microscopy; colocalization (overlay)
appears in yellow and is indicated by white
arrowheads.
O. Maı
¨ga et al. hDlg in ERK cascade
FEBS Journal 278 (2011) 2655–2665 ª2011 The Authors Journal compilation ª2011 FEBS 2657
followed by amplification product sequencing (Fig. S1).
Taken together, the results allow us to conclude that the
larger form of hDlg expressed in hVSMCs corresponds
to the I1A–I1B and I3–I5 insertions, whereas the shorter
form contains I1B and I3–I5 insertions. Both isoforms
contain a Lin-2,-7 domain.
hDlg interacts with MEK2 as assessed by the
yeast two-hybrid system
We then sought to identify hDlg interacting partners
in hVSMCs. Accordingly, we used a vector encoding
the hDlg PDZ1 and PDZ2 domains as bait in a yeast
two-hybrid screening assay of a human aorta cDNA
library. Interestingly, two independent clones were
identified as containing the C-terminal region of the
human MEK2 cDNA. To analyze in more detail the
interacting sites within hDlg and MEK2, mutant deriv-
atives of PGKBT7-PDZ-1-2 and pACT2-MEK2 were
constructed. On the one hand, the conserved GLGF
sequences present in the PDZ1 and PDZ2 domains
were mutated to the positively charged inactive GRRF
sequence [16]. On the other hand, the C-terminal
RTAV putative PDZ-binding motif of MEK2 was
either mutated to RAAV, or deleted. The expression
levels of the mutated forms and of their wild-type
counterparts were similar in yeasts (Fig. 3B, D). These
data suggest that the PDZ1 and PDZ2 domains of
hDlg are separately able to interact with the C-termi-
nus of MEK2, even though the interaction implicating
PDZ2 is stronger, whereas the PDZ3 domain shows
no interaction. Coexpression of the MEK2 mutant
forms with wild-type PDZ-1-2 abolished yeast growth
(Fig. 3C), demonstrating the crucial involvement of the
MEK2 C-terminus in the interaction. Taken together,
these results indicate that the PDZ2 domain of hDlg
and the C-terminal RTAV sequence of MEK2 are
required for the optimal interaction of the two protein
partners.
Coimmunoprecipitation of endogenous hDlg and
MEK2 proteins
To determine whether endogenous hDlg and MEK2
physically interact, coimmunoprecipitation assays were
carried out in HEK-293 (used as a cell model) and in
confluent hVSMC cell lysates. The specific hDlg anti-
body was able to coimmunoprecipate MEK2 from
HEK-293 (Fig. 4A) and hVSMC (Fig. 4C) cell lysates,
whereas, reciprocally, the specific MEK2 antibody
coimmunoprecipitated hDlg from both cell lysates
(Fig. 4B, D). hDlg and MEK2 were not (or minimally)
detectable after immunoprecipitation with irrelevant
antibodies. These results suggest that the hDlg iso-
forms expressed endogenously in HEK-293 cells or in
confluent hVSMCs can physically interact with MEK2,
even though only a small fraction of MEK2 is coim-
munoprecipitated with hDlg, as shown by the large
β1β3β2α12 Ι1Α Ι1Β 3 // 12 13 Ι3 Ι2 Ι5 Ι4 14 15 //
L27A
BC
DE
CXC PDZ1 SH3 GUK
//////
19
//
Primers 1
Primers 2
Primers 3
bp
600
500
Primers 1
Primers 3
Primers 2
Primers GAPDH
132bp
400
300
312
400
300
132300
200
312
Fig. 2. Two hDlg isoforms predominate in human arterial tissues. (A) Schematic representation of the hDlg genomic structure. Open boxes
represent constitutive exons and gray boxes indicate alternatively spliced exons. Three bexons encode an Lin-2,-7 domain (bisoform) and
one aexon a cystein doublet (aisoform). Various combinations of two (I1A and I1B) or four (I2–I5) insertions were described as being tran-
scribed in a tissue-specific manner. Arrows show the relative position of the primer pairs used for RT-PCR. (B–E) Transcripts obtained by
RT-PCR, using (B) primer pair 1, (C) primer pair 2, (D) primer pair 3 or (E) primers directed against GAPDH, on mRNAs extracted from three
different pulmonary artery samples.
hDlg in ERK cascade O. Maı
¨ga et al.
2658 FEBS Journal 278 (2011) 2655–2665 ª2011 The Authors Journal compilation ª2011 FEBS
amount of MEK2 remaining after hDlg precipitation
in hVSMCs (Fig. S2A). Subsequently, the ability of
hDlg to interact with other members of the ERK cas-
cade was tested. Under our experimental conditions,
the hDlg antibody was unable to coimmunoprecipate
MEK1, Raf or ERK1 ⁄2 proteins from hVSMCs
(Fig. S2B).
Colocalization of hDlg and MEK2
The localization of transfected full-length EGFP-hDlg
and human HA-MEK2 in CHO cells was assessed by
confocal immunofluorescence microscopy. EGFP-hDlg
and HA-MEK2 exhibited a diffuse staining with an
occasional patchy appearance and both types of label-
ing partially colocalized in these patches, suggesting
the presence of aggregates (Fig. 5A). In addition, the
localization of endogenous hDlg and MEK2 was
assessed in HEK-293 cells and in hVSMCs. In HEK-
293 cells, hDlg exhibited a general diffuse staining,
although this appeared to be stronger in the region of
the plasma membrane. MEK2 appeared to be more
homogenously distributed in the cytoplasm, although
the two stains overlapped significantly at cell–cell
junctions (Fig. 5B). In hVSMCs, both hDlg and
MEK2 exhibited a diffuse staining, although overlay
revealed a colocalization of the two proteins at some
perinuclear location (Fig. 5C). Taken together, these
results suggest that either transfected or endogenous
hDlg and MEK2 partially colocalize in mammalian
cells.
Discussion
In the present study, we show, for the first time, the
expression of hDlg in the human vascular cell popu-
lation, which is the most abundant in arterial wall
tissue (i.e. the hVSMCs). This protein exists in the
form of two immunoreactive species in membrane
fractions. Subcellular localization experiments suggest
that, in hVSMCs, this MAGUK is mainly associated
with the endoplasmic reticulum, as well as with the
PDZ-1-2
PDZ-1mut-2
PDZ-1-2mut
PDZ3
Media
SD-LT
SD-LT
SD-LT + X-α-gal
RTAV
RAAV
Del
RTAV
RAAV
Del
Ct
50
37
50
Myc
AD
GAPDH
50
AB
CD
Ct
PDZ-1-2
PDZ-1mut-2
PDZ-1-2mu
t
PDZ3
WB
Myc
AD
kDa
50
37
25
50
GAPDH
50
SD-LT + X-α-gal
Fig. 3. The PDZ2 domain of hDlg strongly interacts with the C-terminal sequence RTAV of MEK2. (A, C) Interactions were analyzed by a
yeast two-hybrid assay. (A) The PDZ-1–2 domains of hDlg, either wild-type or mutated on the GLGF sequence in either the PDZ1 domain
(PDZ-1mut-2) or the PDZ2 domain (PDZ-1-2mut), or the hDlg PDZ3 domain, all fused to the GAL4 DNA-binding domain, were co-expressed
in yeasts with the C-terminus of MEK2 fused to the GAL4 activating domain. (C) The PDZ-1-2 domains of hDlg fused to the GAL4 DNA-bind-
ing domain were co-expressed in yeasts with the C-terminus of MEK2, encompassing the PDZ binding sequence, which was intact (RTAV),
replaced by an irrelevant sequence (RAAV), or deleted (Del), all fused to the GAL4 activating domain. Yeasts were grown on two selection
media: SD-LT that selects double transformants, and SD-LTHA + X-a-gal that selects protein–protein interactions with high stringency.
Yeasts grow and turn blue when GAL-4-responsive genes, which encode galactodidases, are activated. (B, D). Western blotting of fusion
protein expression: hDlg PDZ domains fused to the Myc-tagged GAL4 DNA-binding domain were detected by Myc antibody (Myc), and
MEK2 C-terminus fused to the GAL4 activation domain was detected using GAL4 activation domain antibody (AD). Nontransfected yeast
protein extracts were used as control (Ct) and GAPDH detection as a loading control.
O. Maı
¨ga et al. hDlg in ERK cascade
FEBS Journal 278 (2011) 2655–2665 ª2011 The Authors Journal compilation ª2011 FEBS 2659