Báo cáo khoa học: Functional characterization of Drosophila melanogaster PERK eukaryotic initiation factor 2a (eIF2a) kinase pdf

Functional characterization of
Drosophila melanogaster
PERK
eukaryotic initiation factor 2a (eIF2a) kinase
Natalia Pomar, Juan J. Berlanga*, Sonsoles Campuzano, Greco Herna
´
ndez
†
,Mo
´
nica Elı
´
as
and Ce
´
sar de Haro
Centro de Biologı
´
a Molecular ‘Severo Ochoa’, Consejo Superior de Investigaciones Cientı
´
ficas, Universidad Auto
´
noma de Madrid,
Cantoblanco, Madrid, Spain
Four distinct eukaryotic initiation factor 2a (eIF2a)kinases
phosphorylate eIF2a at S51 and regulate protein synthesis in
response to various environmental stresses. These are the
hemin-regulated inhibitor (HRI), the interferon-inducible
dsRNA-dependent kinase (PKR), the endoplasmic reticu-
lum (ER)-resident kinase (PERK) and the GCN2 protein
kinase. Whereas HRI and PKR appear to be restricted to

mammalian cells, GCN2 and PERK seem to be widely
distributed in eukaryotes. In this study, we have character-
ized the second eIF2a kinase found in Drosophila,aPERK
homologue (DPERK).
Expression of DPERK is developmentally regulated.
During embryogenesis, DPERK expression becomes con-
centrated in the endodermal cells of the gut and in the germ
line precursor cells. Recombinant wild-type DPERK, but
not the inactive DPERK-K671R mutant, exhibited an
autokinase activity, specifically phosphorylated Drosophila
eIF2a at S50, and functionally replaced the endogenous
Saccharomyces cerevisiae GCN2. The full length protein,
when expressed in 293T cells, located in the ER-enriched
fraction, and its subcellular localization changed with dele-
tion of different N-terminal fragments. Kinase activity
assays with these DPERK deletion mutants suggested that
DPERK localization facilitates its in vivo function. Similar
to mammalian PERK, DPERK forms oligomers in vivo
and DPERK activity appears to be regulated by ER stress.
Furthermore, the stable complexes between wild-type
DPERK and DPERK-K671R mutant were mediated
through the N terminus of the proteins and exhibited an
in vitro eIF2a kinase activity.
Keywords:eIF2a kinases; Drosophila melanogaster;transla-
tional control; PERK homologue; ER stress.
Protein synthesis is mainly regulated at the initiation of
mRNA translation. Phosphorylation of the a-subunit of
eukaryotic translation initiation factor 2 (eIF2a)isawell
characterized mechanism of translational control (reviewed
in [1,2]). A family of protein kinases phosphorylate eIF2a at

S51 in response to a variety of cellular stresses, including
nutrient starvation, iron deficiency, heat shock, viral infec-
tion and stress signals from the endoplasmic reticulum (ER)
[1,2]. All known eIF2a kinases consist of a conserved
catalytic domain linked to different regulatory regions
which facilitate the different stress signals controlling each
protein kinase. Included in this family are four mammalian
eIF2a kinases: the hemin-regulated inhibitor (HRI), the
double-stranded RNA (dsRNA)-dependent kinase (PKR),
the GCN2 protein kinase and the ER-resident kinase
(PERK, also known as PEK) [1,3]. Additionally, novel
eIF2a kinases from the fission yeast Schizosaccharomyces
pombe [4], and from the malarial parasite Plasmodium
falciparum (PfPK4) [5] have been reported.
The well characterized mammalian eIF2a kinase, HRI, is
expressed most abundantly in erythroid cells [6], although
we found its mRNA and kinase activity in non-erythroid
tissues and in NIH 3T3 cells [7]. HRI becomes activated in
response to heme deficiency and the activity of HRI seems
to be modulated by its association with heat shock proteins
[1,6]. PKR is an interferon-induced dsRNA-activated eIF2a
kinase. It is thought to be activated by dsRNA generated
during viral replication or gene expression [8]. A third
Correspondence to C. de Haro, Centro de Biologı
´
a Molecular
ÔSevero OchoaÕ, CSIC-UAM, Facultad de Ciencias,
Cantoblanco, 28049 Madrid, Spain.
Fax: + 34 91 3974799, Tel.: + 34 91 3978432,
E-mail:

Abbreviations: ATF4, activating transcription factor 4; eIF2a,the
a-subunit (38 kDa) of eukaryotic polypeptide chain initiation fac-
tor 2; HRI, heme regulated inhibitor kinase; PKR, double-stranded
RNA-dependent eIF2a kinase; ER, endoplasmic reticulum;
PERK, PKR-like ER kinase; GCN2, yeast general amino acid control
eIF2a kinase; mHRI, mouse liver HRI; SEK1/2, Schizosaccharomyces
pombe eIF2a kinases; 3-AT, 3-aminotriazole; EST, expressed
sequence tag; ORF, open reading frame; TEMED, N,N,N¢,N¢-
tetramethylethylenediamine; HDM, high density microsomal
fraction; LDM, low density microsomal fraction; SP, signal peptide;
TM, transmembrane; pc, pole cells.
*Present address: Department of Biochemistry, McGill University,
H3G 1Y6 Montreal, Canada.
Present address: Department of Molecular Biology, Gene Expression
Laboratory, Max Planck Institute for Biophysical Chemistry,
Am Fassberg 11, 37077 Go
¨
ttingen, Germany.
Note: The nucleotide sequence(s) data reported in this paper have
been submitted to the EMBL Database and are available under the
accession number(s) AJ 313085.
(Received 6 June 2002, revised 22 October 2002,
accepted 25 November 2002)
Eur. J. Biochem. 270, 293–306 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03383.x
mammalian eIF2a kinase, termed GCN2, was originally
characterized in Saccharomyces cerevisiae as being required
for the amino acid control of GCN4 mRNA translation.
It is activated by uncharged tRNA under amino acid
starvation [9]. The identification of GCN2 homologues
from Drosophila melanogaster [10,11], Neurospora crassa

[12], and mammals [13,14], places GCN2 as one of the best
evolutionarily conserved members of the eIF2a kinase
family [3]. Finally, the mammalian PERK (also known as
PEK) was originally identified in rat pancreatic islet cells
[15]. Mouse PERK is activated by ER stress and contains
a lumenal domain that is similar to the sensor domain of
the ER-stress kinase, Ire1 [16]. PERK homologues have
also been identified in humans and Caenorhabditis elegans
[17]. In addition, sequence analyses have led to the
identification of a putative PERK homologue from
D. melanogaster [17].
In this study, we functionally characterized the Drosophila
PERK eIF2a kinase (DPERK). Northern blot, RT-PCR
and in situ hybridization analyses indicate that DPERK
expression is developmentally regulated. During embryonic
development, DPERK mRNA preferentially accumulates
in the gut endoderm and in the pole cells, after the germinal
band retraction has taken place. Similar to other members
of the eIF2a kinase family, DPERK phosphorylates
Drosophila eIF2a on S50 (S51 in mammals and yeast),
and mediates translational control in yeast. This study
provides evidence of a striking conservation in structure,
function and ER-stress regulation between mammalian and
Drosophila PERK and poses the question of whether
DPERK might be involved, in the same way as mammalian
PERK, in the regulation of gene expression in response to
certain stress signals.
Experimental procedures
Materials
All reagents were from Sigma except ammonium persulfate,

[c-
32
P]ATP and [a-
32
P]dCTP from Amersham Pharmacia
Biotech, and acrylamide, N,N¢-methylenebisacrylamide,
N,N,N¢,N¢-tetramethylethylenediamine (TEMED) and
SDS from Bio-Rad. Specific DNA primers were obtained
from Isogen Bioscience.
Cloning and sequence analysis
The kinase domain sequences of DGCN2 [10] were
compared with the Drosophila General-Bank database
with the aim of finding new eIF2a kinases. All the
sequence analyses were performed using
BLAST
[18],
FASTA
[19],
GAP
(Wisconsin Package, Genetics Computer Group,
University of Wisconsin, Madison) and
CLUSTAL W
[20]
programs. The embryonic expressed sequence tag (EST),
with accession number AA390738, matched as a possible
fragment of a putative PERK-like eIF2a kinase. The EST
AA390738 contained a full length cDNA of 4625 nucle-
otides that hybridizes in Northern blot to a unique
transcript of 4.7 kb. This cDNA displays a large ORF of
3489 bp (nucleotides 282–3770) and a putative polyadeny-

lation signal (AAT AAA, nucleotides 4568–4573). The
5¢-UTR contained three ATG codons followed by
in-frame termination codons, located upstream from the
putative ATG initiation codon (nucleotides 282–284). This
methionine codon is likely to represent the translational
startasitmatchestheDrosophila consensus for transla-
tional initiators [21]. The comparison of the full length
DPERK cDNA with the D. melanogaster genome revealed
that DPERK genomic DNA is encoded within genomic
scaffold 142000013386046 (accession number AE003602)
located in region 83A-83A of chromosome 3R. Recently,
the sequence of DPERK cDNA was published by Sood
et al. [17] and was found to be almost identical to the
DPERK cDNA that we have characterized. Within the
ORF, a single nucleotide change was found in our cDNA
(the triplet CAT, nucleotides 2166–8, is GAT in their
sequence) and the 5¢-and3¢-UTRs in their cDNA are 57
and 337 nucleotides shorter, respectively.
Northern blot analysis
Poly(A)
+
RNA was prepared as described previously [22].
Fifteen lgofPoly(A)
+
RNA from each developmental
stage were separated on a formaldehyde (6%)-agarose (1%)
gel, transferred to a nylon membrane, probed with the full
length DPERK and actin cDNAs radiolabeled with
[a-
32

P]dCTP and analysed by autoradiography.
RT-PCR
Poly(A)
+
RNA was isolated from D. melanogaster
(Oregon R) staged 0–18 h old embryos, 1st, 2nd and 3rd
larvae instars, pupae and adults, by using first, the RNeasy
Mini kit and then the Oligotex mRNA Midi Kit (Qiagen).
Preparations were digested with RNAse-free DNAse I
(Qiagen) to eliminate genomic DNA contamination. cDNA
populations were generated by reverse transcription using
the Marathon cDNA Amplification Kit (Clontech) accord-
ing to the manufacturer’s instructions. Developmental
analysis of mRNA was carried out by PCR using the
Expand-Long Template PCR System (Roche Molecular
Biochemicals) and 6 ng of cDNA from every different
developmental stage as a template under the following
conditions: 94 °C 3 min, 1 cycle, then 94 °C45 s,50°C45 s
and 68 °C 60 s, for 25 or 40 cycles. Primers 5¢-CG
CGAGGAGTACGACTACGATGAGGAAGAG-3¢ and
5¢-CACTGATGCGGCTCACTGGAGCTGCTGAAG-3¢
were used for every amplification experiment to amplify
nucleotides 2646–3778 of the DPERK cDNA. One tenth of
the PCR reaction was loaded on a 1% agarose gel. Primers
5¢-ATGACCATCCGCCCAGCATACAGGCCCAAG-3¢
and 5¢-TGAGAACGCAGGCGACCGTTGGGGTTGG
TG-3¢ were used to amplify nucleotides 1–392 of the
ribosomal protein rp49 ORF under the same conditions to
control the amount of RNA loaded in each lane, as
described previously [23].

Whole-mount embryo RNA
in situ
hybridization
Localization of RNA in whole mount embryos with
antisense digoxigenin-labeled RNA probes was performed
as described [24].
294 N. Pomar et al.(Eur. J. Biochem. 270) Ó FEBS 2003
Antibodies
Based on the DPERK cDNA coding sequence, a syn-
thetic peptide (CG-PKSSGSDDANDDNK) was produced
corresponding to amino acids 873–884 (Fig. 1B), with two
additional residues (CG) at the N-terminal end. The peptide
was synthesized as described by Santoyo et al. [10] and
coupled at the terminal cysteine residue to keyhole limpet
hemocyanin (Calbiochem). Rabbits were immunized as
described by Me
´
ndez and de Haro [25]. For simplicity, the
serum containing anti-DPERK peptide Igs will be referred
to as anti-DPERK Igs. The rabbit polyclonal antibodies
against the ER marker, protein disulfide isomerase (a-PDI)
were the kind gift of J. Gonza
´
lez Castan
˜
o(Universidad
Auto
´
noma de Madrid, Madrid, Spain). The polyclonal
antibody against mannosidase II (Man II) and the mono-

clonal antibody (15C8) against a Golgi integral membrane
protein were kindly provided by G. Egea (Universidad de
Barcelona, Barcelona, Spain) and I. Sandoval (CBMSO,
Madrid, Spain), respectively.
Prokaryotic expression of
Drosophila
eIF2a
The D. melanogaster eIF2a ORF [26] was subcloned into
a pRSETB vector. The pRSETB-eIF2a-S50A mutant
was generated using the QuickChange
TM
Site-Directed
Mutagenesis Kit (Stratagene). Prokaryotic expression was
performed as described previously [7].
Expression of DPERK wild type and mutants in yeast
Nucleotides 274–3767 of DPERK-wt cDNA were ampli-
fied by PCR, together with a V5 tag and a polyhistidine
metal-binding peptide followed by a stop codon in frame
at the C-terminal and introduced into the vector pYX212
(R & D Systems, Inc., Minneapolis, MN, USA).
Fig. 1. The DPERK gene, amino acid sequence
and mutant proteins. (A) Genomic structure of
the DPERK gene. Exons and introns are
shown to scale as boxes and solid lines,
respectively. The coding sequence is indicated
by black boxes. (B) amino acid sequence of the
DPERK protein. Amino acid numbering is
shown on the left. Kinase subdomains are
identified by Roman numerals directly above
the appropriate regions. The predicted signal

peptide (SP) and transmembrane domain
(TM) are indicated by bars above the
sequence. The asterisk denotes the predicted
asparagine-linked glycosylation site. (C)
Schematic diagram of DPERK and DPERK
mutant proteins. The 1162 amino acid-long
wild type DPERK coding sequence is illus-
trated by the larger box. The figures are drawn
to scale. The C-terminal eIF2a kinase domain
contains the 12 catalytic subdomains of Ser/
Thr protein kinases (black boxes), with the
conserved lysine residue (K671) and the insert
region of eIF2a kinases (white box). The reg-
ulatory region (stippled boxes) includes an SP,
TM and the predicted N-linked glycosylation
site (N260). Three deletion mutants are
shown: DSP (in which the first 43 amino acids
containing the signal peptide were deleted);
DTM (in which amino acids 543–569,
containing the transmembrane domain were
deleted); and DNt (in which the first 569 amino
acids containing most of the regulatory
domain were deleted).
Ó FEBS 2003 Molecular characterization of DPERK (Eur. J. Biochem. 270) 295
pYXDPERK-K671R and pYXDPERK-N260A mutants
were created as described above. Deletion mutants were
generated by PCR using primers to produce an ATG
initiation codon in frame, at the following nucleotides of the
DPERK cDNA: DPERK-DSP, nucleotide 410, DPERK-
DNt, nucleotide 2003. DPERK-DTM was generated by PCR

with oligonucleotides that produced a NotI site at nucleotides
1883–2006. All these deletion mutants were also in frame at
the C-terminal end with a V5 tag and a polyhistidine metal-
binding peptide. Yeast GCN2 in pEMBLXyex4 [27] was
kindly provided by C. V. de Aldana.
Plasmids encoding either different DPERK forms or an
empty pYX212 vector were introduced into yeast strains J80
(MATa gcn2D ura3–52 leu2–3 leu2–112 trp1-D63 sui2D
[SUI2-LEU2])andJ82(MATa gcn2D ura3–52 leu2–3
leu2–112 trp1-D63 sui2D [SUI2-S51A LEU2])bytheLiAc
method as described [28]. Transformants were selected by
uracil prototrophy and spotted on to agar plates with
synthetic medium containing 0.67% yeast nitrogen base,
2% glucose and 40 mgÆL
)1
tryptophan (SD) or SD
supplemented with 3-aminotriazole (3-AT) [29]. Agar plates
were incubated for 3 days at 30 °C and photographed.
Yeast extracts
Protein extracts from harvested yeast cells were made by
trichloroacetic acid precipitation after glass bead lysis as
described [30].
Cell cultures and transfections
D. melanogaster Scheneider 2 (S2) cells were maintained in
Complete DES
TM
Expression Medium (Invitrogen) con-
taining 10% (v/v) fetal bovine serum. When specified, S2 cells
were treated with either 0.5, 1 or 2 l
M

thapsigargin (Sigma)
for 80 min or 150, 250 or 500 l
M
dithiothreitol for 5 h. HEK
293T cells were grown in Dulbecco’s modified Eagle’s
medium supplemented with 10% (v/v) fetal bovine serum.
For expression of the fruit fly PERK (DPERK) in S2
cells, the coding sequence from residues 274–3767 was
subcloned into a pMT/V5-His vector (Invitrogen) in frame
with a C-terminal tag encoding the V5 or Myc epitopes
and a polyhistidine metal-binding peptide. Mutants
pMTDPERK-K671R and pMTDPERK-N260A were gen-
erated by introducing a fragment of either pYXDPERK-
K671R or pYXDPERK-N260A containing the appropriate
mutation into pMTDPERK-wt. pMT/V5-His/lacZwas
provided by Invitrogen. Cells were transfected with 19 lgof
plasmid DNA per 35-mm dish using the calcium phosphate
method, as described in the manufacturer’s instructions
(Invitrogen). For cotransfections, the same conditions were
used with 19 lg of plasmid DNA from each construction.
Expression was induced with 500 l
M
copper sulphate for
24 h.
For expression in the mammalian cells, DPERK-wt
and the indicated mutants were subcloned in vectors
pEYFP-N1 (Clontech) or pcDNA3.1 (Invitrogen) in
frame with the YFP signal or V5 epitopes, respectively.
293T cells were plated on 60-mm dishes at 10% conflu-
ence, 12–24 h before transfection. Plasmids (5 lgperdish)

were transfected by the calcium phosphate method, as
described [7].
Immunoprecipitation, eIF2a kinase assay
and immunoblotting
All cells were washed once with NaCl/P
i
(137 m
M
NaCl,
2.6 m
M
KCl, 4 m
M
Na
2
HPO
4
,1.8m
M
KH
2
PO
4
,pH7.4)
and lysed in lysis buffer [20 m
M
Tris/HCl, pH 7.8, 200 m
M
NaCl, 10% (v/v) glycerol, 1% (v/v) Nonidet P-40, 1 m
M

phenylmethylsulfonyl fluoride and a protease inhibitor
cocktail (CompleteÒ, Boehringer Mannheim)]. Cell debris
was removed by centrifugation and the protein concentra-
tion was determined according to Bradford [31]. The
supernatants were subjected to immunoprecipitation or
to SDS/PAGE and blotted onto 0.25 lm nitrocellulose
membranes. Immunoprecipitations were carried out either
with anti-V5 (0.5 lg of Ig2A) (Invitrogen), anti-DPERK
(7 lL of antiserum) or anti-DGCN2 (0.5 lg of affinity-
purified polyclonal antibody) [10] and protein A-Sepharose
with or without competing peptide (5 lg). The immuno-
precipitates were washed twice with lysis buffer, once with
0.5
M
LiCl in NaCl/P
i
and two more times with kinase
buffer (20 m
M
Tris/HCl pH 7.6, 50 m
M
NaCl, 10 m
M
MgCl
2
and 1 m
M
dithiothreitol). The immunoprecipitates
were preincubated for 15 min at 32 °C in the presence of
kinase buffer with 0.1 m

M
ATP and 0.25 mgÆmL
)1
BSA.
All samples were subsequently incubated for 15 min at
32 °C in the presence of recombinant Drosophila eIF2a-wt,
Drosophila eIF2a-S50A or purified rabbit reticulocyte
eIF2 (0.5 lg) as a substrate and 5 lCi of [c
32
P]ATP
(3000 CiÆmmol
)1
), to assay their ability to phosphorylate
eIF2a, as previously reported [10,25]. Incubations were
terminated by addition of SDS sample buffer. Samples
were analysed by electrophoresis on 10% SDS/PAGE,
followed by autoradiography. In order to quantify the
phosphate incorporation into eIF2a, the areas corres-
ponding to the phosphorylated eIF2a were scanned at
633 nm in a computing 300A densitometer (Molecular
Dynamics, Inc.).
The membranes were probed with different antibodies as
indicated in each case: mouse anti-V5 (Invitrogen), mouse
anti-Myc (Invitrogen) or mouse Living Colors Antibody
(Clontech), followed by mouse secondary antibody conju-
gated with horseradish peroxidase. The immunoreactive
bands were detected by enhanced chemiluminiscence (ECL,
Amersham Pharmacia Biotech).
Subcellular fractionation analysis
At 48–72 h post-transfection, 293T cells were washed

twice with NaCl/P
i
and harvested in homogenization
buffer (5 m
M
Hepes/KOH, pH 7.4, 2 m
M
MgCl
2
,1m
M
phenylmethanesulfonyl fluoride and the protease inhibitor
cocktail). After homogenization, using 20 strokes of a
Dounce homogenizer, one volume of a buffer containing
40 m
M
Hepes/KOH, pH 7.4, 2 m
M
EDTA and 0.5
M
sucrose, was added. All operations were performed at
4 °C. The supernatant, taken from a 20-min 19 000 g
centrifugation, was subsequently centrifuged at 45 000 g
for 30 min to obtain a high density microsomal (HDM)
pellet. The HDM pellet was resuspended in HES buffer
(20 m
M
Hepes/KOH, pH 7.4, 1 m
M
EDTA, 0.25

M
sucrose) and the supernatant was centrifuged at
180 000 g for 90 min to obtain a low density microsomal
296 N. Pomar et al.(Eur. J. Biochem. 270) Ó FEBS 2003
(LDM) pellet [32], which was also resuspended in HES
buffer, and a supernatant containing all the soluble
cytoplasm components. Protein concentration in the
different fractions was determined by Bradford analysis
[31] and equivalent amounts of protein were subjected to
SDS/PAGE and immunoblotted using Living Colors
Antibodies.
Results
Molecular characterization of the
D. melanogaster
PERK gene
The discovery, through database searching, of a putative
PERK-like kinase encouraged us to characterize it. The
EST cDNA clone, accession number AA390738, contained
a full-length cDNA. Sequence analysis indicated that the
4625 nucleotide-long Drosophila PERK cDNA contains
281 bp of 5¢ untranslated sequence, 3489 bp of open
reading frame, and 855 bp of 3¢ untranslated sequence
(GenBank accession number AJ313085). The sequence of
DPERK cDNA was compared with the D. melanogaster
genome [33] to verify sequence and determine the genomic
structure of DPERK gene. This analysis revealed the
existence of three introns of 1402, 62 and 84 nucleotides,
respectively, in the DPERK gene (Fig. 1A).
The full-length DPERK cDNA encodes a protein of 1162
amino acids (Fig. 1B), with a predicted molecular mass of

131 kDa as reported previously [17]. The C-terminus eIF2a
kinase domain of DPERK (642–1162) contained all 12
conserved catalytic subdomains of eukaryotic Ser/Thr
protein kinases [34] with an invariant Lys (residue 671) in
subdomain II. The N terminus regulatory domain of
DPERK (residues 1–641) showed characteristic features of
PERK-like kinases [15,16]: the predicted signal peptide (SP,
residues 16–40) and transmembrane (TM, residues 544–563)
domains, obtained through hydropathicity [35] and surface
probability [36] analyses of the protein, and the putative
N-linked glycosylation site (residue 260), given by a
PROSITE
scan of the sequence (www.expasy.ch/scanprosite/) (Fig. 1B).
These data suggested that DPERK may be a glycoprotein
that is not transported to the distal Golgi complex and
resides in the ER.
To understand the role of the N-terminus regulatory
domain of DPERK, we constructed a series of mutants
(Fig. 1C), including a point mutation of the invariant Asn
residue (DPERK-N260A) and deletions that removed: the
signal peptide domain (DPERK-DSP); the entire trans-
membrane domain (DPERK-DTM)ormostoftheregula-
tory domain (DPERK-DNt). The DPERK constructs were
confirmed by sequencing and subcloned into appropriate
expression vectors for functional analysis.
Developmental expression of DPERK
Northern blot analysis of poly(A)
+
RNAs isolated from
different developmental stages revealed a unique DPERK

transcript of approximately 4.7 kb. DPERK is expressed
throughout development, having two major peaks of
expression in early embryo and adult stages (Fig. 2A). Such
a developmental pattern of expression of DPERK was
confirmed by RT-PCR (Fig. 2B).
To determine the pattern of the DPERK mRNA
localization during embryogenesis we performed in situ
hybridization experiments with whole embryos of
D. melanogaster by using a digoxigenin-labeled antisense
RNA probe. DPERK transcripts showed a preferential
accumulation in certain tissues at different developmental
stages (Fig. 3). An even distribution was found in the early
syncytial blastoderm (Fig. 3A). At later blastoderm stages,
the transcripts concentrated in the so-called cortex (Fig. 3B)
and also in the cytoplasm of the cells, save in the pole cells
(pc), the precursors of germ line. At the beginning of the
germ band retraction (stage 11), the DPERK transcripts
accumulated at high levels in the pole cells located at the
posteriormost region of the invaginated midgut rudiment
(Fig. 3C,D). During stage 14 (Fig. 3E,F), and extending
into later stages (Fig. 3G), accumulation of DPERK
transcripts was seen throughout the endodermal cells of
the anterior (amg) and posterior (pmg) midgut. In a dorsal
view, the concentration of the DPERK transcripts in the
gut is more evident (Fig. 3F, inset). The preferential
Fig. 2. Developmental expression of DPERK. (A) Northern blot of
poly(A)
+
RNA prepared from different stages of development was
hybridized with full length DPERK cDNA (top panel). The devel-

opmental stages include embryos (E), first instar larvae (L1), second
instar larvae (L2), pupae (P) and adult flies (A). The arrow indicates
the unique DPERK transcript of 4.7 kb. The filter was rehybridized
with a Drosophila actin probe to control the amount of RNA loaded in
each lane (bottom panel). DPERK transcript levels were much higher
in embryos and adults than in the other stages. (B) RT-PCR analysis
was performed by using poly(A)
+
RNA purified from Drosophila
embryos (E), first instar larvae (L1), second instar larvae (L2), third
instar larvae (L3), pupae (P) and adult (A) stages. Amplification with
the DPERK-specific primers between nucleotides 2646 and 3778 of
DPERK cDNA, as described under ÔExperimental proceduresÕ,
revealed a single fragment of 1.1 kb in each developmental stage (top
panel). Complementary primers to the ribosomal protein rp49 were
used, under same conditions, to control the amount of RNA loaded in
each case (bottom panel). DPERK transcript levels show the same
pattern, confirming the Northern blot analysis.
Ó FEBS 2003 Molecular characterization of DPERK (Eur. J. Biochem. 270) 297
accumulation in the pole cells was maintained at late stage
16, once the pole cells migrated to the gonads (go) (Fig. 3H).
It should be noted that the DPERK expression was not
apparent in the central nervous system where the other
Drosophila eIF2a kinase (DGCN2) was preferentially
expressed [10].
DPERK phosphorylates
Drosophila
eIF2a at S50
in vitro
and mediates translational control in

S. cerevisiae
To verify that DPERK is an eIF2a kinase, we expressed
wild-type DPERK or the presumably inactive DPERK
mutant (DPERK-K671R) in Drosophila S2 cells as
V5-tagged fusion proteins. Recombinant proteins were
immunoprecipitated by using anti-V5 Igs and the immune
complexes were subjected to eIF2a kinase assays as
described under ÔExperimental proceduresÕ. As a positive
control, the reticulocyte heme-reversible HRI [37] was
assayed under the same conditions (Fig. 4A, lanes 1 and
2). DPERK wild-type immune complexes underwent
autophosphorylation (Fig. 4A, lanes 3 and 4) and were
fully active in phosphorylating eIF2a (lane 3). These
phosphorylated DPERK proteins were detected using
immunoblot analysis (Fig. 4A, bottom, lanes 3 and 4).
Moreover, recombinant DPERK specifically phosphoryl-
ated Drosophila eIF2a at S50 (S51 in mammals and yeast)
(Fig. 4A, lane 3) as phosphorylation was not observed in
the assay mixture containing the mutant substrate eIF2a-
S50A (lane 4). Furthermore, we found that the mutant
DPERK-K671R did not phosphorylate itself or eIF2a
(Fig. 4A, lanes 5 and 6), although it was expressed at a
much higher level than wild type DPERK (Fig. 4A,
bottom, lanes 5 and 6). These results are comparable with
those previously observed in both mammalian PERK
[16,17] and mouse HRI [7]. We therefore conclude that
DPERK is, indeed, an eIF2a kinase of D. melanogaster
and has both autokinase and eIF2a kinase activities
in vitro.
It is well known that S. cerevisiae is a useful model

system for studying the in vivo role of eIF2a kinases in
translational control [29]. To address whether DPERK can
functionally replace yeast GCN2, plasmids encoding wild-
type DPERK, the inactive DPERK mutant (DPERK-
K671R) or the vector alone were introduced into two
isogenic yeast strains lacking the GCN2 kinase. Yeast cells
expressing these eIF2a kinases were compared to cells
containing plasmid-encoded yeast GCN2, as a positive
control. As expected, all transformants grew to similar
levels in synthetic medium (Fig. 4B, lower). However only
the cells expressing an active eIF2a kinase, either wild type
DPERK or yeast GCN2, presented normal growth under
amino acid starvation produced by the medium supple-
mented with 3-AT [29] (Fig. 4B, upper). As for the other
eIF2a kinases, the mutation of the invariant lysine in the
kinase subdomain II impaired the ability of DPERK to
rescue growth of yeast lacking GCN2. Moreover, the use
of strain J82 (Dgcn2 SUI2-S51A), isogenic to J80 (gcn2D),
with an alanine substitution for S51 in eIF2a, revealed that
this phosphorylation site was required to provide growth
resistance in the presence of 3-AT (Fig. 4B, upper).
Fig. 3. Expression of DPERK during embry-
onic development. Embryos were hybridized
in situ with antisense DPERK RNA probes.
When appropriate, anterior is left and dorsal is
up. (A) Preblastoderm stage embryo showing
generalized distribution of DPERK mRNA
(probably due to a maternal contribution).
(B) Embryo at the cellular blastoderm stage
presenting generalized distribution of DPERK

mRNA apart from the pole cells (pc).
(C) Lateral view of a stage 11 embryo dis-
playing expression of DPERK almost exclu-
sively at the pole cells (pc). (D) An enlarged
dorsal view of a similar stage 11 embryo
highlights the accumulation of DPERK
mRNA in the pole cells (pc). At progressively
older embryonic stages (E, and F, stage 14; G,
stage 16) expression of DPERK is localized in
the endodermal cells of the anterior (amg) and
posterior (pmg) midgut. This is most easily
visualized in a dorsal view (F and see also the
inset, mg, midgut). (H) Expression of DPERK
persists in the pole cells at late stage 16, once
the pole cells have migrated to the gonads (go).
Hybridization with a sense RNA probe did
not give any appreciable signal (not shown).
Embryonic stages were classified according to
Campos-Ortega and Hartenstein [52].
298 N. Pomar et al.(Eur. J. Biochem. 270) Ó FEBS 2003
Therefore, DPERK, similar to the other eIF2a kinase
family members, requires the regulatory site (S51) in eIF2a
for translational control.
DPERK, but not DGCN2, is activated by ER stress
To date, only two eIF2a kinases have been identified in
D. melanogaster. They are the homologues of yeast GCN2
and mammalian PERK. It is well known that GCN2 is
activated in cells deprived of nutrients [7,9] whereas
mammalian PERK is activated in cells treated with agents
that induce ER stress [16]. We then sought to determine

whether DPERK is specifically activated by some agents
that promote ER stress: thapsigargin and dithiothreitol
[38]. We immunoprecipitated either DPERK (Fig. 5A,D)
or DGCN2 (Fig. 5B,E) from extracts of Drosophila S2
cells treated with increasing concentrations of thapsigargin
(Fig. 5A,B) or dithiothreitol (Figs 5D,E) by using the
appropriate antibodies. The isolated immune complexes
were subjected to eIF2a kinase assays as described under
ÔExperimental proceduresÕ. Immunoprecipitations were
specific because of the fact that they were prevented by
addition of the respective competing peptide immunogen
in the immunoprecipitation assay (Fig. 5A,B,D,E, lanes
2 and 7). We found that both DPERK and DGCN2
immune complexes from untreated S2 cells phosphorylated
the a subunit of eIF2 (Fig. 5A,B,D,E, lane 3), how-
ever, DPERK, but not DGCN2, activity was increased
in cells undergoing ER stress as shown by the in vitro
eIF2a kinase assay of DPERK and DGCN2 immune
complexes (Fig. 5C,F). These results suggest that like
its mammalian homologue, the trigger for Drosophila
DPERK activation is likely to be misfolded proteins in
the ER [16].
Localization and eIF2a kinase activity of DPERK –
the role of the N-terminal domain
It has been proposed that the N-terminal regulatory domain
of mammalian PERK is located in the ER [16]. Because
DPERK also contained a signal peptide at the N terminus
and a hydrophobic domain in the middle of the molecule
(Fig. 1B), we considered the possibility that these similarities
could reflect a similar localization of DPERK.

DPERK-wt and the indicated DPERK deletion mutant
cDNAs specified above (Fig. 1C) were expressed in human
293T fibroblasts as YFP fusion proteins. These DPERK-
transfected cells were subjected to several differential
centrifugations [32] followed by a Western blot analysis of
the enriched high density microsomal (HDM), low density
microsomal (LDM) and cytosolic (CYT) fractions, as
described under ÔExperimental proceduresÕ. Cells expressing
the full length DPERK-wt accumulated the protein in
HDM (Fig. 6A), a fraction where several marker enzyme
activities characteristic of membranes of the ER are mostly
recovered [39]. By contrast, DPERK mutants in which
either the signal peptide (DPERK-DSP) or the entire
transmembrane domain (DPERK-DTM) were deleted,
preferentially accumulated in LDM (Fig. 6A). It is known
that LDM fractions isolated from 3T3-L1 adipocytes using
this procedure mostly contain membranes of the Golgi
apparatus, endosomes and other intracellular membranes
[32]. Finally, a C-terminal YFP-tagged mutant, in which
most of the regulatory domain was deleted (DPERK-DNt)
preferentially accumulated in the cytosolic enriched fraction
(Fig. 6A).
Fig. 4. DPERK is an eIF2a kinase. (A) Autokinase and eIF2a kinase
activities of recombinant DPERK in vitro. Wild type DPERK spe-
cifically phosphorylates the a-subunit of Drosophila eIF2 at residue
S50 (S51 in mammals and yeast). S2 cells were transfected with plas-
mids encoding DPERK-wt or DPERK-K671R containing the V5
epitope. Lysates were subjected to immunoprecipitation with an anti-
V5 Ig, followed by an in vitro phosphorylation assay. These kinase
reactions contained samples of purified HRI from rabbit reticulocyte

lysates as a control (lanes 1 and 2), or either recombinant DPERK-wt
(lanes 3 and 4) or else DPERK-K671R (lanes 5 and 6). The reactions
also included either wild type Drosophila eIF2a (lanes 1, 3 and 5) or
eIF2a-S50A (mut) (lanes 2, 4 and 6). Radiolabeled proteins were
analysed by SDS/PAGE and transferred to an Inmobilon-P mem-
brane followed by autoradiography (top panel). Positions of phos-
phorylated DPERK, HRI, and eIF2a are indicated by arrows. The
same membrane was probed with the anti-V5 Ig in a Western blot
analysis (bottom panel). (B) DPERK functionally substitutes GCN2 in
a yeast model system. Yeast J80 (Dgcn2) and J82 (Dgcn2, SUI2-S51A)
strains were transformed with the indicated eIF2a kinases, or with the
plasmid pYX212 (vector) as a negative control. S. cerevisiae GCN2
was used as a positive control. Patches of transformants were grown in
the appropriate medium as indicated.
Ó FEBS 2003 Molecular characterization of DPERK (Eur. J. Biochem. 270) 299
These studies were further extended to examine the
relative organelle composition of the HDM, LDM and
CYT fractions using antibodies specific to proteins in the
ER and the Golgi apparatus. Thus, in agreement with
recent studies [40], the ER marker protein disulfide
isomerase (a-PDI) was mostly localized in the HDM
fraction (Fig. 6B). In contrast, two distinct markers of the
Golgi complex, a polyclonal antibody against mannosidase
II (Man II) [41] and a monoclonal antibody (15C8), that
recognizes an integral membrane protein located in the cis
and medial Golgi cisternae [42], were found equally
distributed within the HDM and LDM fractions (Fig. 6C).
Previously, very similar results were obtained with fractions
isolated from rat adipose cells [39]. A quantitative approach
reveals that the HDM fraction is enriched in ER, relative to

the other two fractions, but it also contains a significant
Fig. 5. Endogenous DPERK, but not DGCN2, from S2 cells is activated by ER stress. Shown are the results of the in vitro phosphorylation assay of
reticulocyte eIF2 by immune complexes obtained from extracts of S2 cells treated with increasing concentrations of thapsigargin (A and B) or
dithiothreitol (D and E). Cell extracts were subjected to immunoprecipitation with either anti-DPERK (A and D) or anti-DGCN2 (B and E) Igs in
the absence (lanes 3–6) or in the presence (lanes 2 and 7) of the respective competing peptide. Samples of purified HRI from rabbit reticulocytes were
also assayed as a control to position phosphorylated eIF2a (lane 1). The amount of eIF2a phosphorylation was estimated by quantifying the
corresponding band density of the autoradiogram in A and B (see panel C) and in D and E (see panel F). The intensity of the eIF2a band
corresponding to untreated cells (lane 3) was defined as 100%. Similar results were obtained in at least two independent experiments.
300 N. Pomar et al.(Eur. J. Biochem. 270) Ó FEBS 2003
proportion of Golgi membranes. This apparent anomaly
might be due to distinct membrane subspecies of the Golgi
apparatus with sedimentation characteristics similar to
those of the ER membranes.
The results described above, together with preliminary
immunofluorescence analyses of DPERK-wt (data not
shown), strongly suggest that DPERK resides in the ER,
similarly to its mammalian counterpart, and furthermore
that the ER-targeting of DPERK is mediated by these two
unique structural features. To our knowledge, this is the first
report demonstrating a subcellular relocation of this eIF2a
kinase in response to structural changes in the molecule.
It has been proposed that protein targeting plays an
important role in regulating enzymatic activity by providing
access to local substrates or regulatory ligands. Consistent
with the idea that the N-terminus of PERK is important for
mediating activation of this eIF2a kinase, was the previ-
ously reported finding that deletion of these sequences
greatly reduces the catalytic activity of human PERK, but
not that from the C. elegans homologue [17]. In an attempt
to understand the role of the N-terminus domain, as well as

that of the invariant aspargine residue (N260), a predicted
N-linked glycosylation site, in the eIF2a kinase activity of
DPERK, we expressed DPERK wild type and all the
constructed mutants as V5-tagged derivatives, and immu-
noprecipitated them by using anti-V5 antibodies. Because
the expression of the DPERK deletion mutants in S2 cells
was very inefficient, we used HEK 293T cells for expression
of these mutants, under same conditions. These studies
show that all of the immune complexes from either
DPERK-wt or from distinct mutants containing an
unmodified kinase catalytic domain underwent phosphory-
lation and were fully active in phosphorylating eIF2a
(Fig. 7A, top). As shown previously (Fig. 4A), replacing
K671 from DPERK with arginine (K671R) abolished the
ability of the protein to undergo autophosphorylation or to
phosphorylate eIF2a (Fig. 7A, lane 3). We conclude that
the N-terminus of DPERK is not required for in vitro
catalytic activity.
To characterize these DPERK mutants further we tested
whether they would phosphorylate eIF2a at S51 in vivo
and functionally replace yeast GCN2 when expressed in
S. cerevisiae cells, as we previously observed in the wild-type
DPERK (Fig. 4B). All constructs, with the exception of the
plasmid encoding DPERK-DSP, were well expressed in the
J82 strain, however, the expression of the active kinases
(DPERK-wt and the DPERK-N260A mutant) in the
isogenic strain J80 was significantly lower (Fig. 7C). An
inhibition of its own synthesis promoted by the kinase
activity could explain this effect. Even though deletion of the
signal peptide or the transmembrane domain in DPERK

seems not to affect the in vitro eIF2a kinase activity
(Fig. 7A), DPERK mutants in those regions were not able
to maintain eIF2a phosphorylation-dependent cell growth
in yeast (Fig. 7B, upper). Interestingly, all of the mutant
versions of DPERK, that were unable to support yeast
growth under amino acid starvation conditions, were found
to be preferentially accumulated in the LDM fraction
(Fig. 6A), whereas either the wild type or the mutant
DPERK-DNt that showed in vivo catalytic activity were
found to be preferentially accumulated in the HDM or
cytosolic fractions, respectively (Fig. 6A). Altogether, these
Fig. 6. DPERK resides in the endoplasmic reticulum. (A) Distribution
of DPERK-wt and the indicated DPERK mutants in HEK 293T
fibroblasts from subcellular fractionation studies. DPERK-wt and its
mutants were expressed in HEK 293T cells as YFP fusion proteins by
transient transfection of the corresponding plasmids. Microsomal
(HDM, LDM) and cytosolic (CYT) enriched fractions were obtained
from the different lysates as described under ÔExperimental proce-
duresÕ. Protein from cytoplasm or LDM fractions (50 lg), and 20–
30 lg of protein from the HDM fraction were separated on 7.5%
SDS/PAGE and transferred to poly(vinylidene difluoride) membranes
for immunoblotting with anti-YFP Ig. Results are representative of at
least two independent experiments. (B) and (C). Lysates of HEK 293T
fibroblasts were fractionated into cytosol, HDM and LDM as des-
cribed in panel A. Fifty lg of protein from each fraction was resolved
by SDS/PAGE and analysed by immunoblotting using a polyclonal
antibody against protein desulfide isomerase (a-PDI), as a marker of
ER (B) or a monoclonal antibody (15C8) and a polyclonal antibody
against mannosidase II (Man II), as two distinct markers of the Golgi
complex (C). The relative amount of each marker among the subcel-

lular fractions was estimated by quantifying the corresponding band
density of the immunoblots and indicated by the numbers below. The
sum of these relative values was defined as 100. Similar results were
obtained in three independent experiments.
Ó FEBS 2003 Molecular characterization of DPERK (Eur. J. Biochem. 270) 301
results suggest that mislocation of DPERK mutants, rather
than a lack of catalytic activity, prevents eIF2a phosphory-
lation and cell growth in yeast.
The N-terminal regulatory domain is required for the
oligomerization of DPERK
It has been proposed that oligomerization has an important
function in activation of ER stress-signal transducers [43].
Thus, the oligomerization involving the N-terminal ER
lumenal domain is necessary and sufficient to initiate Ire1
activation in yeast [44]. Although Ire1 and PERK share a
weak sequence similarity in their lumenal domains, previous
data suggest that they may use a similar mechanism to sense
ER stress. In fact, treatment of cells with thapsigargin
resulted in the rapid formation of a mammalian PERK-
containing complex [45].
To identify possible in vivo complexes between wild
type DPERK and the inactive DPERK-K671R mutant,
we coexpressed either V5-tagged DPERK-WT or
DPERK-K671R with wild-type or mutant forms of Myc-
tagged derivatives in Drosophila S2 cells, as indicated in
Fig. 7. Expression and activity of DPERK mutants. (A) Autokinase and eIF2a kinase activities of recombinant DPERK mutants in vitro.S2cells
were transfected with plasmids encoding DPERK-wt, DPERK-K671R and DPERK-N260A, while 293T cells were transfected with plasmids
encoding DPERK-DSP, DPERK-DTM, and DPERK-DNt. All the constructs contained the V5 epitope. In vitro kinase reactions contained the
anti-V5 immune complexes prepared from lysates of different transfected cells and purified rabbit reticulocyte eIF2. Purified HRI from rabbit
reticulocyte lysates was included as a control for positioning of phosphorylated eIF2a (lane 1). All of the samples were assayed as described in

Fig. 4A. Thus, after autoradiography (upper panel), the same membrane was probed with monoclonal anti-V5 (lower panel). Positions of
phosphorylated DPERK, HRI and eIF2a are indicated by arrows. Molecular mass markers are indicated on the left. (B) In vivo eIF2a kinase
activity of recombinant DPERK mutants in a yeast model system. Yeast J80 and J82 strains were transformed with high-copy-number plasmids
encoding for the indicated eIF2a kinases, or the plasmid pYX212 (vector) alone. All transformants were analysed as described for Fig. 4B. (C) To
determine the expression of DPERK-wt and its mutants in the J80 and J82 transformants, equal amounts of protein from each cell extract were
resolved by SDS/PAGE and analysed by immunoblot using monoclonal anti-V5 antibodies. Molecular mass markers are indicated on the left.
302 N. Pomar et al.(Eur. J. Biochem. 270) Ó FEBS 2003
Fig. 8A. All constructs were expressed and coexpressed in
S2 cells (Fig. 8A, Input). Importantly, the presence of the
Myc-tagged recombinant proteins in the anti-V5 immune
complexes (Fig. 8A, IP anti-V5, lanes 3–8) under any of the
conditions tested, suggested that DPERK could form a
stable complex in vivo. The isolated immune complexes
containing either WT-WT or WT-K671R oligomers, but
not K671R-K671R oligomers, underwent phosphorylation
and phosphorylated eIF2a in vitro (Fig. 8A, IP kinase,
lanes 5–8). These results indicated that the inactive
DPERK-K671R does not function as a dominant-negative
mutant. Accordingly, the detectable catalytic activity found
in immune complexes obtained from V5-tagged DPERK-
K671R transfected cells (Fig. 8A, lane 2) could be explained
by an interaction between the recombinant mutant protein
and the endogenous DPERK.
To define the role of the N-terminal regulatory domain of
DPERK we performed the same experiment as above, in
lysates of S2 cells cotransfected with Myc-tagged DPERK-
K671R along with the indicated V5-tagged wild type or
mutant forms of DPERK. The recombinant proteins in the
immunoprecipitates were detected by immunoblotting
(Fig. 8A). Also in this case, all constructs were well

expressed in S2 cells with the exception of the plasmid
encoding DPERK-DSP, which was expressed at much
lower levels as previously observed (Fig. 8B, Input). Stable
complexes were found between DPERK-K671R and either
DPERK-N260A or DPERK-DTM (Fig. 8B, IP-antiV5,
lanes 3 and 5) comparable to those observed with wild-type
DPERK (lane 2). Interestingly, the mutant form of DPERK
that lacks a large portion of its N-terminal sequences failed
to associate with DPERK-K671R in an in vivo coimmuno-
precipitation assay (lane 6), even though it was expressed at
much higher amounts than were the other recombinant
proteins. These results suggest that an unknown region in
the N-terminal regulatory domain of DPERK is required
for DPERK oligomerization, whereas the transmembrane
domain, at least, is dispensable.
Discussion
Four distinct protein kinases are known to phosphorylate
eIF2a on S51 in mammals. So far only two of them, the
homologues of GCN2 and PERK (DGCN2 and DPERK),
have been identified in D. melanogaster [10,11,17 and this
report]. Here we show that DPERK can phosphorylate the
regulatory site, S50, in Drosophila eIF2a in vitro and in vivo
and, furthermore, can functionally replace the endogenous
Fig. 8. Oligomerization of DPERK in vivo and eIF2a kinase activity of these oligomers in vitro. (A) Coimmunoprecipitation with anti-V5 Ig of
DPERK-wt and the inactive mutant (K671R) from lysates containing the indicated proteins, tagged with either V5 or Myc epitopes, as shown in the
figure. Immune complexes were subjected to in vitro eIF2a kinase assay (bottom panel, in vitro IP kinase). The same membrane was analysed by a
Western blot using anti-V5 Ig, stripped and immunoblotted against anti-Myc Ig (middle panel, IP anti-V5). The content of DPERK-wt and
DPERK-K671R in the lysates is shown by immunoblotting a sample of the lysate prior to immunoprecipitation (upper panel, Input). Positions of
phosphorylated DPERK and eIF2a are indicated by arrows. (B) Interaction of DPERK-K671R with different DPERK mutants in vivo.Myc-
taggedDPERK-K671R,andV5-taggedDPERKwtandmutantswereusedforimmunoprecipitationwithananti-V5Igandtheimmune

complexes were analysed by Western blot (WB) for the presence of either Myc-tagged or V5-tagged proteins (lower panels, IP anti-V5). As in
Panel A, the content of the indicated proteins in the lysates is shown by immunoblotting a sample of the lysate prior to immunoprecipitation (upper
panels, Input).
Ó FEBS 2003 Molecular characterization of DPERK (Eur. J. Biochem. 270) 303
yeast GCN2 kinase by a process requiring the S51
phosphorylation site in yeast eIF2a. These results allow
unambiguous classification of DPERK as an eIF2a kinase.
Drosophila DPERK is 49% identical to rat PERK within
the catalytic domain and its amino-terminal regulatory
domain shares two unique structural features with mam-
malian PERKs, a signal peptide and a transmembrane
region. Additionally, the N terminus of DPERK contains
the predicted N-linked glycosylation site (N260), conserved
in mammalian PERK and Ire1 [16]. As we found that the
N260 residue is not required for catalytic activity of
DPERK, we do not know the functional significance of
this invariant residue yet.
It was reported previously that mammalian PERK is a
transmembrane protein resident in the ER membrane
whose activity is repressed by the ER chaperone BiP.
When too many unfolded proteins accumulate in the ER,
BiP dissociates from PERK, resulting in the activation of
this kinase [44,45]. We have shown here that recombinant
DPERK, when expressed in 293T cells, also localized in
the membranes of the ER. It is interesting to note that the
two highly hydrophobic regions of this eIF2a kinase are
required for its appropriate subcellular localization of
DPERK. Thus, in the absence of each one of these two
sequences, the majority of the respective DPERK mutants
moved from ER to the Golgi apparatus fractions, whereas

the mutant that essentially contains the kinase catalytic
domain showed a cytoplasmic localization. It is especially
noteworthy that while all of these DPERK deletion
mutants showed an in vitro eIF2a kinase activity, only
the full-length protein and the C-terminal derivative,
resident in the ER membrane and the cytoplasm, respect-
ively, can function in translational control in the yeast
model. We speculate that the subcellular localization of
DPERK facilitates its in vivo function in the sense that a
localization in either the rough ER or the cytoplasm,
where the concentration of ribosomes is particularly high,
would facilitate the accessibility of DPERK to its substrate
and therefore its kinase function in vivo. In this respect,
sequence motifs have been identified in PKR and GCN2
that are required for ribosomal association [9,46]. Because
DPERK does not possess these motifs, its possible
association with the ribosome could be established by this
alternative mechanism.
As mentioned before, mammalian cells have at least
four eIF2a kinases, and each one is activated by different
signals. Thus, PERK is responsible for phosphorylation of
eIF2a when mammalian cells are subjected to agents that
induce ER stress (thapsigargin, tunicamycin and dithio-
threitol), whereas GCN2 is activated by amino acid or
serum deprivation [13,16,38]. In this report we have
shown that S2 cells treated with thapsigargin or dithio-
threitol specifically increased the endogeneous DPERK
kinase activity, whereas the activity of endogenous
DGCN2 was not affected. These similarities between
mammalian and Drosophila PERK indicate that both

proteins may use a similar mechanism to transduce ER
stress.
It has been proposed that oligomerization has an
important function in the activation of ER stress-signal
transducers [44,45]. Hence, oligomerization is sufficient to
activate the kinase activity of mammalian PERK upon
treatment of cells with thapsigargin [45]. Other eIF2a
kinases are also activated by oligomerization. Thus, recent
results have shown that the regulatory domain of yeast
GCN2, related to histidyl-tRNA synthetase, contains a
dimerization domain that is required for tRNA binding and
kinase activation [47]. Furthermore, the binding of dsRNA
to the regulatory domain of PKR is required for dimeriza-
tion and activation of this kinase in vivo [48]. The data
reported here indicate that recombinant DPERK shows the
ability to form oligomers when expressed in S2 cells, and
that its N-terminal regulatory domain is required for
DPERK oligomerization. In good agreement with these
novel data, it has been recently reported that the N-terminal
lumenal domain of human Ire1a forms stable dimers [49],
and furthermore, that a region in the ER lumenal domain
mediates PERK oligomerization in mammalian cells [50]. It
is possible that, like other eIF2a kinases, DPERK oligome-
rization mediated by its regulatory domain is both necessary
and sufficient for activation of its eIF2a kinase activity
in vivo.
DPERK is dynamically expressed during embryogene-
sis. At later stages, DPERK expression concentrates in the
gut and the gonads. Interestingly, the highest levels of
mammalian PERK mRNA expression were also found in

secretory tissues [17]. On the other hand, note that at the
same developmental stages, DGCN2 expression is restric-
ted to a few cells of the central nervous system [10] where
DPERK mRNA was not detected. This surprising
selectivity, together with its highly regulated expression,
is consistent with the idea that Drosophila eIF2a kinases
might be involved in determining cell identity and
underscores their role in translational control during
Drosophila development.
It is well known that phosphorylation of eIF2a by
yeast GCN2 kinase mediates gene-specific translational
control of GCN4 in S. cerevisiae. The GCN4 gene
encodes a transcription factor that controls the expression
of genes involved in amino acid biosynthesis, and GCN2
activity is required for increased translation of GCN4
mRNA and for cell survival during amino acid starvation
[9]. Recent results have shown that this gene-specific
regulation also occurs in higher eukaryotes in response to
modest levels of eIF2a phosphorylation. Thus, both
mammalian GCN2 and PEKR, when activated by their
cognate upstream stress signals, repressed translation of
most mRNAs but selectively increased translation of
Activating Transcription Factor 4 (ATF4), resulting in
the induction of a downstream target gene, CHOP [38].
The striking conservation in structure and function
between mammalian and Drosophila eIF2a kinases
prompted us to speculate whether the above mechanism
for transcriptional activation of gene expression might
also be conserved. In this respect, the Drosophila homo-
logue of the mammalian ATF4, the cryptocephal gene,

controls molting and metamorphosis in Drosophila [51].
Therefore, it will be of great interest to determine whether
the Drosophila eIF2a kinases are involved in the trans-
lational control of mRNAs that encode key growth
regulating proteins. We speculate that the role of DPERK
in cellular physiology will be important in the control of
cell growth, differentiation and, thus, development in
Drosophila.
304 N. Pomar et al.(Eur. J. Biochem. 270) Ó FEBS 2003
Acknowledgements
This work was supported in part by DGICYT Grant PM98-0128 (to C.
de H) and by an institutional grant from the Fundacio
´
nRamo
´
n Areces
(to the Centro de Biologı
´
a Molecular ÔSevero OchoaÕ). N.P. is the
recipient of a predoctoral fellowship from the Fundacio
´
nRamo
´
n
Areces (Spain) and J.J.B. is the recipient of a postdoctoral fellowship
from the Comunidad de Madrid (Spain).
We gratefully acknowledge Javier Santoyo’s preliminary work in the
database search and thank Ignacio V. Sandoval and Vassiliki Lalioti
for their assistance with subcellular fractionation analyses. We are
grateful to Encarnacio

´
nMartı
´
nez-Salas for comments on the manu-
script. We also thank Jose
´
Alcalde for excellent technical assistance.
References
1. de Haro, C., Me
´
ndez, R. & Santoyo, J. (1996) The eIF2a kinases
and the control of protein synthesis. FASEB J. 10, 1378–1387.
2. Mathews, M.B., Sonenberg, N. & Hershey, J.W.B. (2000) Origins
and principles of translational control. In Translational Control of
Gene Expression (Sonenberg, N., Hershey, J.W.B. & Mathews,
M.B., eds), pp. 1–31. Cold Spring Harbor Laboratory Press, Cold
Spring Harbor, NY, USA.
3. Dever, T.E. (1999) Translation initiation: adept at adapting.
Trends Biochem. Sci. 24, 398–403.
4. Zhan, K., Vattem, K.M., Bauer, B.N., Dever, T.E., Chen, J.J. &
Wek, R.C. (2002) Phosphorylation of eukaryotic initiation factor
2 by heme-regulated inhibitor kinase-related protein kinases
in Schizosaccharomyces pombe is important for resistence to
environmental stresses. Mol. Cell. Biol. 22, 7134–7146.
5. Mo
¨
hrle, J.J., Zhao, Y., Wernli, B., Franklin, R.M. & Kappes, B.
(1997) Molecular cloning, characterization and localization of
PfPK4, an eIF2a kinase-related enzyme from the malarial parasite
Plasmodium falciparum. Biochem. J. 328, 677–687.

6. Chen, J.J. (2000) Heme-regulated eIF2a kinase. In Translational
Control of Gene Expression (Sonenberg,N.,Hershey,J.W.B.&
Mathews, M.B., eds), pp. 529–546. Cold Spring. Harbor Labor-
atory Press, Cold Spring Harbor, NY, USA.
7. Berlanga, J.J., Herrero, S. & de Haro, C. (1998) Characterization
of the hemin-sensitive eukaryotic initiation factor-2a kinase from
mouse nonerythroid cells. J. Biol. Chem. 273, 32340–32346.
8. Kaufman, R.J. (2000) The double-stranded RNA-activated pro-
tein kinase PKR. In Translational Control of Gene Expression
(Sonenberg,N.,Hershey,J.W.B.&Mathews,M.B.,eds),pp.503–
527. Cold Spring Harbor Laboratory Press, Cold Spring Harbor,
NY, USA.
9. Hinnebusch, A.G. (1997) Translational regulation of yeast GCN4.
A window on factors that control initiator-tRNA binding to the
ribosome. J. Biol. Chem. 272, 21661–21664.
10. Santoyo, J., Alcalde, J., Mendez, R., Pulido, D. & de Haro, C.
(1997) Cloning and characterization of a cDNA encoding a pro-
tein synthesis initiation factor-2a (eIF-2a)kinasefromDrosophila
melanogaster. Homology to yeast GCN2 protein kinase. J. Biol.
Chem. 272, 12544–12550.
11. Olsen, D.S., Jordan, B., Chen, D., Wek, R.C. & Cavener, D.R.
(1998) Isolation of the gene encoding the Drosophila melanogaster
homolog of the Saccharomyces cerevisiae GCN2 eIF2a kinase.
Genetics 149, 1495–1509.
12. Sattlegger, E., Hinnebusch, A.G. & Barthelmess, I.B. (1998) cpc-3,
the Neurospora crassa homologue of yeast GCN2, encodes a
polypeptide with juxtaposed eIF2a kinase and histidyl-tRNA
synthetase-related domains required for general amino acid con-
trol. J. Biol. Chem. 273, 20404–20416.
13. Berlanga, J.J., Santoyo, J. & de Haro, C. (1999) Characterization

of a mammalian homolog of the GCN2 eukaryotic initiation
factor-2a kinase. Eur. J. Biochem. 265, 754–762.
14. Sood, R., Porter, A.C., Olsen, D.A., Cavener, D.R. & Wek, R.C.
(2000) A mammalian homologue of GCN2 protein kinase
important for translational control by phosphorylation of
eukaryotic factor-2a. Genetics 154, 787–801.
15. Shi, Y., Vattem, K.M., Sood, R., An, J., Liang, J., Stramm, L. &
Wek, R.C. (1998) Identification and characterization of pancreatic
eukaryotic initiation factor-2a subunit kinase, PEK, involved in
translational control. Mol. Cell. Biol. 18, 7499–7509.
16. Harding, H.P., Zhang, Y. & Ron, D. (1999) Protein translation
and folding are coupled by an endoplasmic-reticulum-resident
kinase. Nature 397, 271–274.
17. Sood, R., Porter, A.C., Ma, K., Quilliam, L.A. & Wek, R.C.
(2000) Pancreatic eukaryotic initiation factor-2a kinase (PEK)
homologues in humans. Drosophila melanogaster and Caenor-
habditis elegans that mediate translational control in response to
endoplasmic reticulum stress. Biochem. J. 346, 281–293.
18. Altschul, S.F., Gish, W., Miller, W., Myers, E.W. & Lipman, D.J.
(1990) Basic local alignment search tool. J. Mol. Biol. 215, 403–
410.
19. Pearson, W.R. & Lipman, D.J. (1988) Improved tools for biolo-
gical sequence comparison. Proc. Natl Acad. Sci. USA 85, 2444–
2448.
20. Thompson, J.D., Higgins, D.G. & Gibson, T.J. (1994) CLUSTAL
W: improving the sensitivity of progressive multiple sequence
alignment through sequence weighting, position-specific gap pe-
nalties and weight matrix choice. Nucleic Acids Res. 22, 4673–
4680.
21. Cavener, D.R. & Ray, S.C. (1991) Eukaryotic start and stop

translation sites. Nucleic Acids Res. 19, 3185–3192.
22. Campuzano, S., Carramolino, L., Cabrera, C.V., Ruiz-Go
´
mez,
M., Villares, R., Boronat, A. & Modolell, J. (1985) Molecular
genetics of the achaete-scute gene complex of D. melanogaster. Cell
40, 327–338.
23. Al-Atia, G.R., Fruscoloni, P. & Jacobs-Lorena, M. (1985)
Translational regulation of mRNAs for ribosomal proteins during
early Drosophila development. Biochemistry 24, 5798–5803.
24. Jiang, J., Kosman, D., Ip, Y.T. & Levine, M. (1991) The dorsal
morphogen gradient regulates the mesoderm determinant twist in
early Drosophila embryos. Genes Dev. 5, 1881–1891.
25. Me
´
ndez, R. & de Haro, C. (1994) Casein kinase II is implicated
in the regulation of heme-controlled translational inhibitor of
reticulocyte lysates. J. Biol. Chem. 269, 6170–6176.
26. Qu, S. & Cavener, D.R. (1994) Isolation and characterization of
the Drosophila melanogaster eIF-2a gene encoding the alpha
subunit of translation initiation factor eIF-2. Gene 140, 239–
242.
27. Cesareni, G. & Murray, J.A.H. (1987) In Genetic Engineering:
Principals and Methods (Setlow, J.K. & Hollaender, A., eds), Vol.
9, pp. 135–154. Plenum Press, New York, NY, USA.
28. Schiestl, R.H. & Gietz, R.D. (1989) High efficiency transformation
of intact yeast cells using single stranded nucleic acids as a carrier.
Curr. Genet. 16, 339–346.
29. Dever, T.E. (1997) Using GCN4 as a reporter of eIF2a phos-
phorylation and translational regulation in yeast. Methods 11,

403–417.
30. Foiani, M., Marini, F., Gamba, D., Lucchini, G. & Plevani, P.
(1994) The B subunit of the DNA polymerase a-primase com-
plex in Saccharomyces cerevisiae executes an essential function at
the initial stage of DNA replication. Mol. Cell. Biol. 14, 923–
933.
31. Bradford, M.M. (1976) A rapid and sensitive method for the
quantification of microgram quantities of protein utilizing the
principle of protein-dye binding. Anal. Biochem. 72, 248–254.
32. Piper, R.C., Hess, L.J. & James, D.E. (1991) Differential sorting
of two glucose transporters expressed in insulin-sensitive cells.
Am. J. Physiol. 260, C570–C580.
Ó FEBS 2003 Molecular characterization of DPERK (Eur. J. Biochem. 270) 305
33. Adams, M.D. et al. (2000) The genome sequence of Drosophila
melanogaster. Science 287, 2185–2195.
34. Hanks, S.K. & Hunter, T. (1995) The eukaryotic protein kinase
superfamily: kinase (catalytic) domain structure and classification.
FASEB J. 9, 576–596.
35. Kyte, J. & Doolittle, R.F. (1982) A simple method for displaying
the hydrophatic character of a protein. J. Mol. Biol. 157, 105–132.
36. Emini, E.A., Hughes, J.V., Perlow, D.S. & Boger, J. (1985)
Induction of hepatitis A virus-neutralizing antibody by a virus-
specific synthetic peptide. J. Virol. 55, 836–839.
37. Me
´
ndez, R., Moreno, A. & de Haro, C. (1992) Regulation of
heme-controlled eukaryotic polypeptide chain initiation factor-2a
subunit kinase of reticulocyte lysates. J. Biol. Chem. 267, 11500–
11507.
38. Harding, H.P., Novoa, I., Zhang, Y., Zeng, H., Wek, R.,

Schapira, M. & Ron, D. (2000) Regulated translation initiation
controls stress-induced gene expression in mammalian cells. Mol.
Cell 6, 1099–1108.
39. Simpson, I.A., Yver, D.R., Hissin, P.J., Wardzala, L.J., Karnieli,
E., Salans, L.B. & Cushman, S.W. (1983) Insulin-stimulated
translocation of glucose transporters in the isolated rat adipose
cells: characterization of subcellular fractions. Biochim. Biophys.
Acta 763, 393–407.
40. Lalioti, V.S., Vergarajauregui, S., Pulido, D. & Sandoval, I.V.
(2002) The insulin-sensitive glucose transporter, GLUT4, interacts
physically with Daxx. J. Biol. Chem. 277, 19783–19791.
41. Luna, A., Matas, O.B., Martı
´
nez-Mena
´
rguez, J.A., Mato, E.,
Dura
´
n, J.M., Ballesta, J., Way, M. & Egea, G. (2002) Regulation
of protein transport from the Golgi complex to the endoplasmic
reticulum by CDC 42 and N-WASP. Mol. Biol. Cell. 13, 866–879.
42. Yuan, L., Barriocanal, J.G., Bonifacino, J.S. & Sandoval,
I.V. (1987) Two integral membrane proteins located in the cis-
middle and trans-part of the Golgi system acquire sialylated N-
linked carbohydrates and display different turnovers and sensi-
tivity to cAMP-dependent phosphorylation. J. Cell Biol. 105, 215–
227.
43. Sidrauski, C., Chapman, R. & Walter, P. (1998) The unfolded
protein response: an intracellular signalling pathway with many
surprising features. Trends Cell Biol. 8, 245–249.

44. Liu, C.Y., Schro
¨
der, M. & Kaufman, R.J. (2000) Ligand-
independent dimerization activates the stress response kinase IRE
1 and PERK in the lumen of the endoplasmic reticulum. J. Biol.
Chem. 275, 24881–24885.
45.Bertolotti,A.,Zhang,Y.,Hendershot,L.M.,Harding,H.P.&
Ron, D. (2000) Dynamic interaction of BiP and ER stress trans-
ducers in the unfolded-protein response. Nature Cell Biol. 2, 326–
332.
46. Zhu, S., Romano, P.R. & Wek, R.C. (1997) Ribosome targeting of
PKR is mediated by two double-stranded RNA-binding domains
and facilitates in vivo phosphorylation of eukaryotic initiation
factor-2. J. Biol. Chem. 272, 14434–14441.
47. Qiu, H., Dong, J., Hu, C., Francklyn, C.S. & Hinnebusch, A.G.
(2001) The tRNA-binding moiety in GCN2 contains a dimeri-
zation domain that interacts with the kinase domain and is
required for tRNA binding and kinase activation. EMBO J. 20,
1425–1438.
48. Zhang, F., Romano, P.R., Nagamura-Inoue, T., Tian, B., Dever,
T.E., Mathews, M.B., Ozato, K. & Hinnebusch, A.G. (2001)
Binding of double-stranded RNA to protein kinase PKR is
required for dimerization and promotes critical autopho-
sphorylation events in the activation loop. J. Biol. Chem. 276,
24946–24958.
49. Liu, C.Y., Wong, H.N., Schauerte, J.A. & Kaufman, R.J. (2002)
The protein kinase/endoribonuclease Ire1a that signals the
unfolded protein response has a luminal N-terminal ligand-
independent dimerization domain. J. Biol. Chem. 277, 18346–
18356.

50. Ma, K., Vattem, K.M. & Wek, R.C. (2002) Dimerization and
release of molecular chaperone inhibition facilitate activation of
eukaryotic initiation factor-2 kinase in response to endoplasmic
reticulum stress. J. Biol. Chem. 277, 18728–18735.
51. Hewes, R.S., Schaefer, A.M. & Taghert, P.H. (2000) The cryp-
tocephal gene (ATF4) encodes multiple basic-leucine zipper pro-
teins controlling molting and metamorphosis in Drosophila.
Genetics 155, 1711–1723.
52. Campos-Ortega, J.A. & Hartenstein, V. (1985) In The Embryonic
Development of Drosophila Melanogaster (Campos-Ortega, J.A. &
Hartenstein, V. eds), pp. 3–84. Springer-Verlag, Berlin, Germany.
306 N. Pomar et al.(Eur. J. Biochem. 270) Ó FEBS 2003