Occupancy of the catalytic site of the PDE4A4 cyclic AMP phosphodiesterase by rolipram triggers the...
-
Upload
robert-terry -
Category
Documents
-
view
214 -
download
0
Transcript of Occupancy of the catalytic site of the PDE4A4 cyclic AMP phosphodiesterase by rolipram triggers the...
www.elsevier.com/locate/cellsig
Cellular Signalling 15 (2003) 955–971
Occupancy of the catalytic site of the PDE4A4 cyclic AMP
phosphodiesterase by rolipram triggers the dynamic redistribution of this
specific isoform in living cells through a cyclic AMP independent process
Robert Terrya,*, York-Fong Cheungb, Morten Praestegaarda, George S. Baillieb,Elaine Hustonb, Irene Gallb, David R. Adamsc, Miles D. Houslayb
aBioImage A/S, Moerkhoej Bygade 28, Soeborg DK-2860, DenmarkbMolecular Pharmacology Group, Division of Biochemistry and Molecular Biology, Wolfson Building, Institute of Biomedical and Life Sciences,
University of Glasgow, Glasgow G12 8QQ, Scotland, UKcDepartment of Chemistry, Heriot-Watt University, Riccarton, Edinburgh EH14 4AS, Scotland, UK
Received 21 March 2003; accepted 30 May 2003
Abstract
In cells transfected to express wild-type PDE4A4 cAMP phosphodiesterase (PDE), the PDE4 selective inhibitor rolipram caused PDE4A4
to relocalise so as to form accretion foci. This process was followed in detail in living cells using a PDE4A4 chimera formed with Green
Fluorescent Protein (GFP). The same pattern of behaviour was also seen in chimeras of PDE4A4 formed with various proteins and peptides,
including LimK, RhoC, FRB and the V5-6His tag. Maximal PDE4A4 foci formation, occurred over a period of about 10 h, was dose-
dependent on rolipram and was reversible upon washout of rolipram. Inhibition of protein synthesis, using cycloheximide, but not PKA
activity with H89, inhibited foci generation. Foci formation was elicited by Ro20–1724 and RS25344 but not by either ArifloR or RP73401,
showing that not all PDE4 selective inhibitors had this effect. ArifloR and RP73401 dose-dependently antagonised rolipram-induced foci
formation and dispersed rolipram pre-formed foci as did the adenylyl cyclase activator, forskolin. Foci formation showed specificity for
PDE4A4 and its rodent homologue, PDE4A5, as it was not triggered in living cells expressing the PDE4B2, PDE4C2, PDE4D3 and PDE4D5
isoforms as GFP chimeras. Altered foci formation was seen in the Db-LR2-PDE4A4 construct, which deleted a region within LRZ, showing
that appropriate linkage between the N-terminal portion of PDE4A4 and the catalytic unit of PDE4A4 was needed for foci formation. Certain
single point mutations within the PDE4A4 catalytic site (His505Asn, His506Asn and Val475Asp) were shown to ablate foci formation but
still allow rolipram inhibition of PDE4A4 catalytic activity. We suggest that the binding of certain, but not all, PDE4 selective inhibitors to
PDE4A4 induces a conformational change in this isoform by ‘inside-out’ signalling that causes it to redistribute in the cell. Displacing foci-
forming inhibitors with either cAMP or inhibitors that do not form foci can antagonise this effect. Specificity of this effect for PDE4A4 and
its homologue PDE4A5 suggests that interplay between the catalytic site and the unique N-terminal region of these isoforms is required.
Thus, certain PDE4 selective inhibitors may exert effects on PDE4A4 that extend beyond simple catalytic inhibition. These require protein
synthesis and may lead to redistribution of PDE4A4 and any associated proteins. Foci formation of PDE4A4 may be of use in probing for
conformational changes in this isoform and for sub-categorising PDE4 selective inhibitors.
D 2003 Elsevier Inc. All rights reserved.
Keywords: PDE4A4; Rolipram; Cyclic AMP phosphodiesterase; Green flourescent protein; Protein kinase A
0898-6568/03/$ - see front matter D 2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0898-6568(03)00092-5
Abbreviations: PDE, cyclic nucleotide phosphodiesterase; PDE4,
cAMP specific family 4 PDE; UCR, Upstream Conserved Region;
PDE4A4, human PDE4 long form known specifically as HSPDE4A4B
(GenBank Accession number, L20965); IBMX, isobutylmethylxanthine;
rolipram, 4-[3-(cyclopentoxyl)-4-methoxyphenyl]-2-pyrrolidone (racemic
mixture used unless specified otherwise); IC50, concentration of inhibitor
yielding 50% inhibition; EC50, concentration of compound yielding 50%
stimulatory effect; LR2, Linker Region 2.
* Corresponding author. Tel.: +45-4443-7506.
E-mail address: [email protected] (R. Terry).
1. Introduction
Cyclic AMP controls a wide variety of cellular functions
[1]. As the only means of degrading this second messenger
is through the action of cyclic nucleotide phosphodiesterases
(PDEs), these enzymes provide a key regulatory system. Over
30 different PDEs, able to degrade cAMP in mammalian
cells, have been identified [1–5]. These are expressed in a
cell-type specific fashion, implying distinct functional roles.
R. Terry et al. / Cellular Signalling 15 (2003) 955–971956
Signalling through cAMP is compartmentalised [1,6,7],
with gradients of cAMP identified in living cells and the
pivotal role that PDEs play in determining these gradients
demonstrated [8–10]. Undoubtedly the specific intracellular
targeting that characterises many PDE isoforms plays a key
role in establishing compartmentalised cAMP signalling [3].
There is currently considerable interest [11–17] in PDE4
cAMP specific phosphodiesterases as PDE4 selective inhib-
itors have potent anti-inflammatory action and are being
developed as therapeutic agents for respiratory diseases.
However, side effects, such as emesis, which are associated
with certain PDE4 inhibitors such as the archetypal PDE4
selective inhibitor rolipram, have hindered their therapeutic
deployment [11,12]. PDE4 enzymes are encoded by four
genes (A, B, C, D) [2,3,17]. Each PDE4 gene, as a
consequence of alternative mRNA splicing, generates mul-
tiple PDE4 isoforms, which are each characterised by a
unique N-terminal region. These various isoforms are then
grouped into so-called ‘long’, ‘short’ and ‘super-short’
variants, dependent upon either the presence or absence of
regulatory UCR1 and UCR2 modules. Various PDE4 iso-
forms have been shown to interact with other proteins and
lipids, allowing specific isoforms to be targeted to distinct
intracellular sites and signalling complexes within cells
[3,17]. Of key importance in achieving such recruitment
are the isoform-specific unique N-terminal regions, which
was first shown for the PDE4A1 isoform [18]. Thus, for
example, the RACK1 [19] and AKAP signalling scaffold
proteins [20,21] can recruit PDE4D5 and PDE4D3, respec-
tively, and the TAPAS-1 domain of PDE4A1 can insert into
phospholipid bilayers [22].
Chimeras generated with Green Fluorescent Protein
(GFP) have been used extensively to monitor the intracel-
lular targeting and dynamics of a wide variety of proteins in
living cells [23]. Here we exploit this technology to identify
a novel action associated with the PDE4 selective inhibitor,
rolipram, namely its ability to cause the intracellular redis-
tribution of the PDE4A4 long isoform in living cells. We
evaluate this action to show that it is specific for PDE4A4
and that not all PDE4 inhibitors can trigger such a response.
This leads us to propose that rolipram and certain other
PDE4 selective inhibitors may exert actions consequent
upon binding to PDE4A4 that are not mediated by cAMP.
2. Materials and methods
[3H]-cyclic AMP and ECL reagent were from Amersham
International (Amersham, UK). Dithiothreitol, Triton X-100
and N-{1-(2,3-dioleoyloxy)propyl}-N,N,N-trimethylammo-
nium methylsulfate (DOTAP) and protease inhibitor tablets
were obtained from Boehringer Mannheim (Mannheim,
Germany). Bradford reagent was from Bio-Rad (Herts,
UK). All other biochemicals were from Sigma (Poole,
UK) with (R)-(�)- and (S)-(+)-enantiomers of rolipram from
BioMol (Pennsylvania, USA).
2.1. Molecular biology constructs
Analyses were done on the long PDE4A4B isoform
(GenBank accession number L20965) and a chimera formed
by making an in-frame fusion of PDE4A4B linked at its C-
terminus to Green Fluorescent Protein (GFP; GenBank
accession number U55672). Similar C-terminus chimeras
were made to PDE4A4B using, in place of GFP, human
LimK-1, RhoC, the FRB fragment of the FKBP12 rapamycin
associated protein, FRAP (FRB= FRAP[2025–2114]) and
the minimal V5-6His tag peptide (V5 epitope plus 6 histi-
dines). The latter, derived from the pEF6/V5-His C cloning
vector (Invitrogen) adds 47 amino acids to PDE4A4B. Site
directed mutagenesis was performed using a QuickChange
DNA mutagenesis kit (Stratagene, La Jolla, CA, USA)
according to the manufacturer’s instructions. All mutagenesis
and deletion constructs were confirmed by DNA sequencing.
The cDNA clones and expression vectors used for the various
PDE4 isoforms have been described previously [24,25].
2.2. Generation of stable cells lines
Chinese hamster ovary cells (CHO), human HEK293 and
rat RBL cells were transfected with the plasmids described
using the transfection agent FuGENEk 6 (Boehringer
Mannheim, USA) according to the method recommended
by the suppliers. Stable transfectants were selected using 1
mg/ml G418 sulphate (Geneticin; Invitrogen, Denmark) in
the growth medium most appropriate to each cell type. CHO
cells were cultured in HAM’s F12 nutrient mix with
Glutamax-1, 10% foetal bovine serum (FBS) plus 0.1%
penicillin/streptomycin (10000 units ml� 1). For the culture
of HEK293 and RBL cells, Dulbecco’s Modified Eagles
Medium (DMEM) containing 4500 mg/l glucose and Glu-
tamax was used in place of HAM’s F12. All cells were
cultured at 37 jC in 95% humidity and conditions of normal
atmospheric gases supplemented with 5% CO2. Clonal cell
lines were established from populations of stably transfected
cells by transferring single cells to individual culture com-
partments on Lab-Tek chambered coverglasses (Nunc, Den-
mark) containing fresh culture medium with 1 mg/ml G418
sulphate.
2.3. Analysis of GFP-PDE4 chimera in living cells
Cells were cultured to about 80% confluence in Lab-Tek
chambered coverglasses. Cells were also grown in plastic
96-well plates (Polyfiltronics Packard 96-View Plate or
Costar Special Optics Plate; both types tissue culture trea-
ted). Prior to experiments, the cells were cultured over night
in the appropriate medium, without G418 sulphate but with
100 Ag/ml penicillin–streptomycin mixture and 10% FBS.
Confocal images were collected using a Zeiss LSM 410
microscope (Carl Zeiss, Jena, Germany) equipped with
argon-ion and HeNe lasers giving excitation lines at 488
and 546 nm, respectively. GFP fluorescence was collected at
R. Terry et al. / Cellular Signalling 15 (2003) 955–971 957
488-nm excitation using a FT510 dichroic beam splitter and
a 510- to 525-nm band pass emission filter. Red fluoro-
phores (Texas Red, Alexa Fluor 546, BODIPY 558/568,
etc.) were collected at a 546-nm excitation using a FT560
dichroic beam splitter and an OG590 long-pass emission
filter. Images were typically collected with a Fluar 40X, NA:
1.3 oil immersion objective, the microscope’s confocal
aperture set to a value of 10 units (optimum for this lens).
Image sequences of live cells over time, used to monitor the
time course for the formation of accretion foci and their
behaviour during cell division, were collected using a Zeiss
Axiovert 135M fluorescence microscope fitted with a Fluar
40X, NA: 1.3 oil immersion objective and coupled to an
Orca ER charged coupled device (CCD) camera (Hamma-
matsu Photonics, Japan). Filters used for GFP were a 470F20-nm excitation filter, a 510-nm dichroic mirror and a
515F 15-nm emission filter for minimal image background.
The cells were maintained at 37 jC with a custom-built stage
heater, and the cell chambers were sealed with high-vacuum
grease to prevent significant pH shifts in the full culture
media used for these time lapse experiments. Image analysis
methods were used to measure the number of accretion foci
per cell. Cells were generally fixed and their nuclei stained
prior to imaging using: 4% formaldehyde in phosphate
buffered saline (PBS, pH7.4; InVitrogen, Denmark) for 15
min, followed by 10 AM Hoechst 33258 (Molecular Probes,
Eugene, OR, USA) in PBS for a further 10 min, then washed
three times in PBS.
Automated images were collected on a Nikon Diaphot
300 (Nikon, Japan) using a Nikon Plan Fluor 20X/0.5NA
objective lens. The microscope was fitted with a motorised
specimen stage and motorised focus control (Prior Scientific,
Cambridge UK), excitation filter wheel (Sutter Instruments,
Novato, CA, USA) and Orca ER camera. Automation of
stage positioning, focus, excitation filter selection and image
acquisition was controlled using macros written in-house,
running under IPLab for Windows (Scanalytics, Fairfax, VA,
USA). Images were collected in pairs, the first using a 340/
10-nm excitation filter, the second with a 475RDF40 exci-
tation filter (Chroma, Brattleboro, VT). Both images were
collected via the same dichroic and emission filters opti-
mised for EGFP applications (XF100 filter set, Omega
Optical, Brattleboro, VT). Image pairs were automatically
analysed using custom macros written in IPLab software,
involving for each image pair the identification and counting
of individual nuclei from the Hoechst image, then identifi-
cation and counting of accretion foci from the corresponding
GFP image. The ratio of foci number to nuclear count for
each pair represents an estimate of the average number of
accretion foci per cell in that image pair. The analysis also
returned the total area and intensity of pixels in identified
accretion foci, allowing these parameters to be appreciated
as average values per cell. The amount of GFP in cells was
also quantified using a fluorescence plate reader optimised
for reading fluorescence from adherent cells (Fluoroskan
Ascent CF, Labsystems, Finland) equipped with appropriate
filter sets (EGFP: excitation 485 nm, emission 527 nm;
Hoechst 33258: excitation 355 nm, emission 460 nm). For
such measurements, it was usual to increase the cell density
to approximately 1�105 cells/well (200 Al/well) to obtain
best signal to background ratio. Total GFP signal/well was
measured first from live cells in full growth medium, then
cells were treated with an extraction buffer to simultaneous-
ly fix the cells and extract the mobile GFP, leaving immobile
GFP for subsequent measurement. Extraction buffer con-
sisted of 0.4% formaldehyde buffer (pH 7.4) plus 1% Triton
X-100. Extraction buffer was applied to cells for 10 min at
room temperature. Full fixation and nuclear staining was
completed with 4% formaldehyde buffer + 10 AM Hoechst
33258 for a further 10 min, then cells were washed three
times in PBS to complete the procedure. Once the cells had
been extracted, fixed and their nuclei stained with Hoechst
dye, the immobile GFP signal was measured and corrected
per well for cell number using the Hoechst signal. All plate
reader data have been corrected for relevant background
offsets.
2.4. Antibodies
Antisera specific for the human PDE4A sub-family have
been described elsewhere by us [26]. Briefly, this was
generated against a fusion protein formed between GST
and a portion of the extreme C-terminal region of PDE4A
that is unique to this sub-family. Such a region is found in
common to all active PDE4 isoforms. Attempts to character-
ise components potentially associated with the PDE4A4
accretion foci included use of specific antisera raised against
the following proteins: pericentrin (Covance); h-tubulin and
g-tubulin (both from Sigma); Human transferrin receptor
(Zymed); Fyn (15) and Lyn (44) (both from Santa Cruz
Biotech); PKA-RI, PKA-RII, PKA-RII, AKAP121,
AKAP149, AKAP KL, AKAP82, AKAP220 and AKAP450
were all from BD Transduction Labs. Secondary antibodies
to detect these antisera were obtained from Molecular
Probes, and were generally conjugated to Alexa Fluor 546
or similar fluorophore with fluorescence characteristics
distinct from GFP. BODIPY 558/568 Phalloidin (Molecular
Probes) was used to identify F-actin structures in cells.
2.5. Transient expression in COS7 cells
As described previously by us [24], transfection was done
using the COS7 SV40-transformed monkey kidney cell line
maintained at 37 jC in an atmosphere of 5% CO2/95% air in
complete growth medium containing DMEM supplemented
with 0.1% penicillin/streptomycin (10000 units ml� 1), glu-
tamine (2 mM) and 10% FCS. Briefly, COS7 cells were
transfected using DEAE-dextran. The DNA to be transfected
(5 Ag) was mixed, and incubated for 15 min with 250 Al of 10mg ml� 1 DEAE-dextran in PBS to give a ‘DNA-dextran’
mix. When cells reached 70% confluency, in 100-mm dishes,
the medium was removed and the cells were given 10 ml of
R. Terry et al. / Cellular Signalling 15 (2003) 955–971958
fresh DMEM containing 0.1 mM chloroquine and the DNA-
dextran mix (250 Al). The cells were then incubated for 4 h at37 jC. After this period, the medium was removed and the
cells shocked with 10% DMSO in PBS. After PBS washing,
the cells were returned to normal growth medium and left for
a further 2 days before use. For determination of PDE
activity, the cells were homogenized in KHEM buffer (50
mM KCl, 10 mM EGTA, 1.92 mM MgCl2, 1 mM dithio-
threitol, 50 mM Hepes, final pH7.2,) containing ‘complete’
protease inhibitors (Boehringer Mannheim) of final concen-
trations 40 Ag/ml PMSF, 156 Ag/ml benzamine, 1 Ag/ml
aprotonin, 1 Ag/ml leupeptin, 1 Ag/ml pepstatin A and 1 Ag/ml antipain. As described previously, in such transfected
cells then >98% of the total PDE activity was due to the
recombinant PDE4 isoenzyme.
2.6. Assay of cAMP PDE activity
PDE activity was determined [25] by a modification of
the two-step procedure of Thompson and Appleman [27] in
10 mM MgCl2, 20 mM Tris–HCl buffer final pH7.4. All
assays were conducted at 30 jC with initial rates taken from
Fig. 1. Rolipram causes the recruitment of PDE4A4–GFP into accretion foci in
fluorescence images of PDE4A4 chimeras in living cells. CHO cells stably transf
after having been treated with rolipram (3 AM) for 7 h. HEK cells stably transfe
having been treated with rolipram (10 AM) for 16 h. RBL cells (e) stably transfec
AM) for 24 h. In image (f) is shown a confocal fluorescent image of CHO cells tran
for 18 h prior to fixation and subsequent visualisation of PDE4A4 using a PDE4A4
images from experiments repeated at least three times.
linear time courses. Activity was linear with added protein
concentration. Dose-effect inhibitor analyses were per-
formed with 1 AM cAMP as a substrate over a range of
rolipram concentrations. Rolipram was dissolved in 100%
DMSO as a 10-mM stock and diluted in 20 mM Tris–HCl
(pH 7.4), 10 mM MgCl2 buffer to provide a range of
concentrations for use in the assay. The residual levels of
DMSO were not found to affect PDE activity over the
ranges used in this study. Inhibitor studies were analysed
using the Kaleida-Graph software package (Synergy Soft-
ware, Reading, PA). The activities of the various PDE4A4
mutant forms are expressed relative (%) to that of wild-type
PDE4A4 by assaying, at a substrate concentration of 1 AMcAMP, equal amounts of immunoreactive PDE4A4 protein,
as determined with PDE4A-specific antibodies in both
ELISA and quantitative immunoblotting [25].
2.7. SDS/PAGE and Western blotting
Eight-percent acrylamide gels were used and the samples
boiled for 5 min after being resuspended in Laemmli buffer
[28]. Gels were run at 8 mA/gel overnight or 50 mA/gel for
living CHO and HEK cells. Images (a) through (e) show confocal GFP
ected to express PDE4A4–GFP in (a) the absence of rolipram and also (b)
cted to express PDE4A4–GFP in (c) the absence of rolipram and (d) after
ted to express PDE4A4–GFP shown here after treatment with rolipram (10
sfected to express untagged PDE4A4 and challenged with rolipram (10 AM)
-specific antibody. White arrows indicate examples of foci. These are typical
R. Terry et al. / Cellular Signalling 15 (2003) 955–971 959
4–5 h with cooling. For detection of transfected PDE by
Western blotting, 2–50 Ag protein samples were separated
by SDS/PAGE and then transferred to nitrocellulose before
being immunoblotted using the indicated specific antisera.
Labelled bands were identified using peroxidase linked to
anti-rabbit IgG, and the Amersham ECLWestern blotting kit
was used as a visualisation protocol.
2.8. Protein analysis
Protein concentration was determined using BSA as
standard [29].
3. Results
3.1. Rolipram causes the redistribution of PDE4A4–GFP in
living cells
PDE4A4 [30], having both UCR1 and UCR2, is a so-
called PDE4 long isoform [2,3,17]. It has been found in
brain and also in various cell types associated with immune
responses [2,3,31–33]. It is characterised by its unique N-
terminal region of 107 amino acids [30], which can func-
tionally interact with the SH3 domains of various SRC
family tyrosyl kinases [25,26].
Fig. 2. Rolipram causes the recruitment of PDE4A4 tagged with various different p
confocal fluorescence images of the various indicated PDE4A4 chimeras in fixed
transfected to express PDE4A4-LimK-1 (full length, 647 residues; a,b), PDE4A4-R
residues; e, f). Shown are untreated cells (a, c, e) and cells challenged with rolipram
HEK (h) cells that have been transiently transfected with PDE4A4 tagged at its C-te
rolipram (10 AM) for 18 h prior to fixation. White arrows indicate examples of fo
Here we show a fluorescence confocal microscopy
analysis of living CHO (Fig. 1a) and HEK (Fig. 1c) cells
that have been stably transfected so as to express a chimera
formed from PDE4A4 and Green Fluorescent Protein
(PDE4A4–GFP). While fluorescence is dispersed through
the cell interior, it is clearly excluded from nuclei and
certain other structures in the cytoplasm of these cells and
which, in terms of disposition and appearance, are most
consistent with mitochondrial and endoplasmic reticulum
compartments. Strikingly, however, the addition of the
PDE4 selective inhibitor, rolipram (10 AM), causes a pro-
found change in the distribution of PDE4A4–GFP in both
CHO (Fig. 1b) and HEK (Fig. 1d) cells which took the form
of the appearance of accretion foci. Such foci are also
generated by rolipram treatment of RBL monocytic cells
expressing PDE4A4–GFP (Fig. 1e). However, they are not
seen upon rolipram treatment of cells transfected to express
GFP by itself, where fluorescence remains spread uniformly
throughout the cell, including the nucleus (data not shown).
Rolipram-induced foci are also evident for untagged
PDE4A4 expressed in CHO cells treated with rolipram
(Fig. 1f), as well as with PDE4A4 C-terminally tagged with
the V5 epitope plus six histidine residues (V5-6His chimera)
expressed in both CHO and HEK cells (Fig. 2g,h). PDE4A4
fusions to a range of very different proteins also respond to
rolipram in the same way as PDE4A4–GFP. LimK-1 (full-
roteins into accretion foci in living CHO cells. Images (a) through (h) show
cells visualised using a PDE4A4-specific antibody. CHO cells were stably
hoC (full length, 193 residues; c, d) and PDE4A4-FRB (FRAP fragment, 89
(10 AM) 18 h prior to fixation. Also shown are examples of both CHO (g) and
rminus with the minimal V5-6His peptide and that had been challenged with
ci. These are typical images from experiments repeated at least three times.
R. Terry et al. / Cellular Signalling 15 (2003) 955–971960
length, 647 residues; Fig. 2a,b), RhoC (full-length, 193
residues; Fig. 2c,d) and FRB (FRAP fragment, 89 residues;
Fig. 2e,f) each fused to PDE4A4 in place of GFP success-
fully form accretion foci when challenged with rolipram.
Foci are clearly separate from centrioles, as detected by
immunostaining for pericentrin and from the microtubule
motor centre, as indicated by staining for gamma-tubulin
(data not shown). They are also separate from the cellular
actin network and from the anchored form of PKA, as
detected using antisera to the RIIa and RIIh subunits of
PKA (data not shown).
The other major PDE4 isoforms that are widely expressed
in cells, including those of the immune system, are the
PDE4B2 short isoform and the PDE4D3 and PDE4D5 long
isoforms. We show here, however, that GFP chimeras made
with these species and expressed in CHO cells do not
redistribute to form accretion foci in the presence of rolipram
(Fig. 3a–c). Neither does the PDE4C2 long isoform (data
not shown). However, rolipram-induced accretion foci were
seen in CHO cells stably expressing a GFP chimera formed
Fig. 3. Analysing the potential of rolipram to alter the intracellular distribution of
confocal, widefield fluorescence images of PDE4-GFP chimeras in living cells.
PDE4D5-GFP, (d) PDE4A5-GFP were all challenged with rolipram (10 AM) for
arrows in (d) indicate the only examples of foci formed by rolipram in these studie
with PDE4A5 (Fig. 3d), the rodent homologue of the human
PDE4A4 isoform.
3.2. Properties of rolipram-induced redistribution of
PDE4A4–GFP in living cells
We show here the time dependence of the ability of 10
AM rolipram to cause the dynamic redistribution of
PDE4A4–GFP in living cells over a period of 9 h (Fig.
4a). During this time, the total GFP fluorescence, expressed
per cell, increased by 52F 3%. The increase in total GFP
fluorescence is very likely to be due in part to cAMP-
mediated stimulation of the hCMV promoter used to drive
expression of the PDE4A4–GFP chimera, as shown previ-
ously by others [34,35]. However, strikingly, the immobile
fraction of fluorescence, reflecting PDE4A4–GFP in accre-
tion foci, increased by more than 10-fold during the 9-
h period of treatment with 10 AM rolipram (Fig. 4a). The
formation of accretion foci is completely reversible (Fig. 4b)
as removal of rolipram leads to complete dispersal of foci.
chimera of various other PDE4-GFP isoforms in living CHO cells. Shows
CHO cells transfected to express (a) PDE4B2-GFP, (b) PDE4D3-GFP, (c)
18 h prior to visualisation of GFP fluorescence in these living cells. White
s. Typical experiments of the ones done at least three times are shown here.
Fig. 4. The rolipram-induced formation accretion foci in PDE4A4–GFP
expressing CHO cells. (a) Shows the change over time in both the total
PDE4A4–GFP signal and also the signal due to the immobilised
PDE4A4–GFP localised in accretion foci in CHO cells treated with
rolipram (10 AM). Values are meanF S.D. (n= 8); (b) CHO cells expressing
PDE4A4–GFP were treated for 16 h with rolipram (10 AM) so as to allow
the formation and assessment of accretion foci (time = zero). They were
then washed and incubated in fresh medium without rolipram for 4.5 h
before re-assessment in order to define total PDE4A4–GFP fluorescence
and the immobile GFP fraction remaining after the removal of rolipram.
During this period of rolipram-free incubation, total GFP content of the
cells decreased by approximately 17%. Values are meanF S.D. (n= 32).
R. Terry et al. / Cellular Signalling 15 (2003) 955–971 961
As an example, within 4.25 h after washout of rolipram, all
accretion foci disperse and the immobile PDE4A4–GFP
signal accordingly falls dramatically to approximately 10%
of the value measured in cells that have not had rolipram
removed; during the same time period, these washout cells
lose only 17F 3% of their total PDE4A4–GFP (Fig. 4b).
Interestingly, reappearance of foci in cells that had been
previously challenged with rolipram occurs at a consider-
ably quicker rate than in naı̈ve cells challenged with
rolipram for the first time (Fig. 5a,b). This indicates that
rolipram sensitises the cell in some fashion and that this
‘memory’ extends past the time taken for dispersal of foci
by washout of rolipram. In this regard, we see here that the
addition of cycloheximide (30 Ag/ml) to PDE4A4–GFP
expressing CHO cells challenged for 16 h with a range of
rolipram concentrations up to 100 AM completely ablated
foci formation (Fig. 6a). These data indicate that additional
protein synthesis is needed for rolipram-induced foci for-
mation to ensue despite the presence of high levels of
PDE4A4–GFP in these cells.
Rolipram elicits PDE4A4–GFP foci generation in a
dose-dependent fashion (Fig. 6a), with an EC50 value of
0.50F 0.08 AM (meanF S.D.; n = 4 experiments). This
value is similar to that which we observe for the dose-
dependent inhibition of PDE4A4–GFP activity by rolipram
(Table 1) and to that reported previously by us [25] for the
inhibition of wild-type PDE4A4 by rolipram.
Additionally, challenge with the adenylyl cyclase activa-
tor, forskolin (up to 100 AM), either alone or together with
the nonselective PDE inhibitor IBMX (100 AM), failed to
induce foci formation in PDE4A4–GFP expressing CHO
cells (data not shown). Thus, foci generation is unlikely to
be due to any elevation in intracellular cAMP levels. Indeed,
challenging cells with forskolin actually serves to inhibit
foci generation dose-dependently, and will disperse foci pre-
formed with rolipram (Fig. 6b; 5 AM) with an EC50 value of
0.44F 0.04 AM (meanF S.D., n = 3) in the continued
presence of rolipram (5 AM). These data suggest that
interaction of the reversible competitive inhibitor rolipram
with PDE4A4 engenders a conformational change in this
isoform that directly leads to generation of accretion foci,
and that elevated cAMP levels may reverse this effect by
competing with rolipram for interaction with the active site
of PDE4A4.
3.3. PDE4 selective inhibitors can be segregated into those
that elicit foci formation of PDE4A4–GFP in living cells
and those that do not
We set out to evaluate whether PDE4 selective inhibitors
(Fig. 7a), other than rolipram, might elicit the generation of
foci in PDE4A4–GFP expressing CHO cells. Like rolipram,
the PDE4 selective inhibitors Ro20–1724 and RS25344
[36–38] both cause the dose-dependent formation of foci in
PDE4A4–GFP expressing cells (Fig. 7b), with EC50 values
of 3.8F 0.2 AM and 17F 3 nM, respectively (meanF S.D.;
n = 3 experiments). These values are very similar to the IC50
values of 4.3F 0.5 AM and 22F 7 nM we note here for the
inhibition of PDE4A4–GFP by Ro20–1724 and RS25344,
respectively (meanF S.D.; n = 3 experiments). Intriguingly,
(S)-(+)-rolipram is much less effective than its (R)-(�)-
enantiomer (Fig. 7b). This is with regard to both the
magnitude of foci generation, with (S)-(+)-rolipram gener-
ating maximum levels of foci around 20% of that of (R)-
(�)-rolipram, and the dose-dependency of its action, with
Fig. 5. Rolipram causes the formation of accretion foci more rapidly in PDE4A4–GFP expressing CHO cells that have previously been subjected to challenge
with rolipram. In (a) are shown CHO cells expressing PDE4A4–GFP with images taken at the indicated times after the addition of rolipram (10 AM). In the
first panel of images (i, ii, iii), cells had not previously been exposed to rolipram. In the second panel of images (iv, v, vi), cells were pretreated with rolipram
(10 AM) for 16 h so as to allow the formation and assessment of accretion foci. They were then washed and incubated in fresh medium without rolipram for 4.5
h at 37 jC so that all rolipram-induced foci of PDE4A4–GFP had dispersed. Cells were then returned to incubator conditions between images; (b) open circles
show the time dependence of PDE4A4–GFP foci reappearance in cells that have been pretreated for 16 h with rolipram (10 AM), then washed to allow foci to
disperse and then, finally, rechallenged (zero time) with rolipram (10 AM). The second curve (filled circles) shows the time dependence of PDE4A4–GFP foci
formation in naive cells exposed to rolipram (10 AM) for the first time. Both curves report foci number per cell vs. time. Values for the reappearance curves
were derived from single fields of more than 240 cells followed over the period shown. All images used in this analysis were collected using the same
microscope settings and counts made using a fixed threshold for spot recognition. Values have been normalised to values from cells exposed to rolipram (10
AM) for 16 h, taken, for this purpose, as the maximum response of the system. The de novo foci formation curve uses values for the immobile GFP fluorescence
(cell corrected) derived from wells treated for various times with rolipram (10 AM). The values for this curve are given as a percentage of the immobile GFP
fluorescence from cells treated for 25 h with rolipram (10 AM), taken again as the maximal response of the system.
R. Terry et al. / Cellular Signalling 15 (2003) 955–971962
Fig. 6. Formation and dispersal properties of rolipram-induced PDE4A4–
GFP accretion foci in CHO cells. (a) Shows the dose dependence for the
formation of accretion foci (immobile GFP per cell) in response to rolipram
alone (filled circles) and in the presence of 30 Ag/ml cycloheximide (open
circles); in each case, PDE4A4–GFP expressing CHO cells were incubated
with the indicated concentrations of rolipramF 30 Ag/ml cycloheximide for
16 h. Each value is the meanF S.D. (n= 8). (b) Shows the dose dependence
for the forskolin-mediated dispersal of rolipram-induced formation of
accretion foci of PDE4A4–GFP in CHO cells. Cells treated with 5 AMrolipram for 18 h, then in continuous presence of 5 AM rolipram, exposed to
a range of forskolin concentrations + 100 AM IBMX (all treatments) for a
period of 2 h. Values are meanF S.D. (n= 8).
Table 1
Inhibition of PDE4A4 constructs by rolipram
PDE4A4–GFP
construct
IC50 (AM) for
inhibition by
rolipram
EC50 (AM) for
foci formation
by rolipram
EC50 (mM)
for activation
by Mg2 +
Relative
activity
(%)
Wild-type 4A4 0.83F 0.06 0.50F 0.08 1.5F 0.5 (100)
H506N-4A4 0.77F 0.13 Not formed 12F 2 48F 8
H505N-4A4 0.27F 0.08 Not formed 4F 1 48F 4
V475D-4A4 0.63F 0.15 Not formed 17F 3 23F 4
This table shows the concentration of rolipram (IC50; AM) giving 50%
inhibition of the indicated recombinant PDE4A4–GFP constructs assayed
in the presence of 1 AM cAMP as substrate (n= 4 separate determinations).
Shows the concentration of Mg2 + (EC50; mM) giving 50% activation of the
indicated recombinant PDE4A4–GFP constructs assayed in the presence of
1 AM cAMP as substrate (n = 3 separate determinations). PDE4A4
unconjugated had an EC50 value of 2F 1 mM Mg2 +. Also shown is the
relative activity (%) that the mutant forms have with respect to the wild-
type enzyme (100%), when assayed in the presence of 1 AM cAMP as
substrate. (n= 6, separate determinations) Values are given as means with
errors as S.D.
R. Terry et al. / Cellular Signalling 15 (2003) 955–971 963
(S)-(+)-rolipram exhibiting an EC50 value of 3.9F 0.9 AMcompared to that of 0.48F 0.09 AM, for (R)-(�)-rolipram
(meanF S.D.; n = 4 experiments).
Remarkably, in contrast to the action of rolipram and
these two other PDE4 selective inhibitors, another pair of
PDE4 selective inhibitors, namely ArifloR [11,39,40] and
RP73401 [41], are completely unable to cause foci forma-
tion in PDE4A4–GFP (data not shown). Indeed, as with
forskolin (Fig. 6c), both ArifloR and RP73401 are able to
antagonise the formation of foci by rolipram and actually to
disperse foci pre-formed by rolipram (Fig. 7c). This occurs
in a dose-dependent fashion with IC50 values of 2.6F 0.4
AM for ArifloR and 15F 2 nM for RP73401 when 5 AMrolipram is employed to generate foci (meanF S.D.; n = 3
experiment values). Such values reflect the different affin-
ities of these two inhibitor compounds for inhibiting the
catalytic activity of PDE4A4 [25].
We observed that treatment of CHO cells with either
ArifloR or RP73401 causes an increase in total GFP
fluorescence to a similar extent to that seen using rolipram.
Thus, for example, we show here the dose-dependent
increase in the total cellular fluorescence from cells express-
ing the PDE4A4–GFP chimera treated with either RP73401
or rolipram (Fig. 8a). In both cases, the maximum level of
increase of the total GFP signal is 32F 7% over a 16-
h incubation period (Fig. 8a), although at no concentration
of RP73401 did we observe any accretion foci produced.
This indicates that any increase in the amount of cellular
PDE4A4–GFP caused by the effects of treatment with
certain PDE4 inhibitors does not necessarily translate into
foci formation. In other words, formation of accretion foci
with rolipram is not simply a cellular response to increased
levels of PDE4A4 product, as it is not seen with RP73401.
Additionally, the facile reversibility of foci formation, by
simply washing out rolipram, under conditions where little
or no change in total PDE4A4–GFP levels occur, also
militates against foci generation being due to inappropriate
processing.
We also record a further difference in the action of these
two PDE4 selective inhibitors. This arose from experiments
where we first treated cells with either rolipram or RP73401
for 10 h, washed them free of inhibitor and then left them in
inhibitor-free medium for 2.5 h to allow the foci to disappear
in the rolipram-treated cells. After this, we challenged both
groups of cells with 5 AM rolipram and followed the rate of
(re)formation of accretion foci. While this rate was very
R. Terry et al. / Cellular Signalling 15 (2003) 955–971964
much faster in the rolipram-pretreated cells, the RP73401-
pretreated cells respond with a rate of foci formation that was
equivalent to that of naı̈ve cells (Fig. 8b). Thus, while both
rolipram and RP73401 cause a similar increase in the total
GFP signal over 10 h, through increasing the expression
levels of PDE4A4–GFP, it is only pretreatment with roli-
pram that serves to sensitize the cells to rolipram-induced
foci formation. This indicates that sensitisation is not due to
any general PDE4 inhibitor-induced increase in PDE4A4–
GFP expression, which is presumably mediated through an
increase in intracellular cyclic AMP. Instead, sensitisation is
specifically associated with certain PDE4 selective inhibi-
tors, such as rolipram, which are able to elicit a discrete sig-
nalling event consequent upon binding to PDE4A4.
3.4. Generation of PDE4A4 mutants that are resistant to
foci formation by rolipram
It has been suggested that PDE4 isoforms can exist in
distinct conformational states [13,37,42] that can be dis-
criminated by their differential sensitivity to inhibition by
the (R)-(�) and (S)-(+) enantiomers of rolipram; specifically
a heightened sensitivity to the (R)-(�), compared to (S)-(+),
enantiomer. The novel process that we have identified here,
involving PDE4A4, clearly shows discrimination in the
ability to effect foci generation by these two enantiomers
of rolipram (Fig. 7b). We suggest that the binding of
rolipram to the PDE4A4 catalytic site may engender a con-
formational change that is required for the accretion of
PDE4A4 to foci. Such a condition implies that it might be
possible to employ mutagenesis to effect alterations in the
catalytic site of PDE4A4 that serve to ablate the accretion of
PDE4A4 to foci. An exhaustive mutagenesis procedure is
not tenable, however. This is not only because the catalytic
unit is large, but also because functional assessment of mu-
tants requires that clonal cell lines stably expressing them be
generated, each of which constitutes a very considerable
undertaking. Nevertheless, we set out to try and address the
principle that foci formation results from a distinct confor-
mational change consequent on the binding of rolipram to
the active site rather than by serving merely to effect in-
hibition of cAMP hydrolysis. To do this, we studied muta-
tion of a few key residues, which rational consideration of
the enzyme’s structure suggested might play a role in the
proposed conformational change induced by rolipram.
Although no crystal structure is currently available for
any full-length PDE4 enzyme, that of the PDE4B catalytic
unit has been derived [43]. As this region is highly con-
served (>90%) throughout the PDE4 family [3], the
PDE4B2 catalytic unit structure is likely to provide a good
Fig. 7. The action of various other PDE4 selective inhibitors on the rolipram
induced accretion foci of PDE4A4–GFP in CHO cells. (a) Structures of the
indicated PDE4 selective inhibitors; (b) dose-effect analyses of the
formation of accretion foci of PDE4A4–GFP in CHO cells treated with
the indicated compounds Ro20–1724, RS25344, (R)-(�)-rolipram and (S)-
(+)-rolipram. MeanF S.D. (n= 4); (c) dose-effect analyses of the inhibition
of formation of rolipram-induced accretion foci of PDE4A4–GFP in CHO
cells treated with the indicated concentrations of either ArifloR (open
circles) or RP73401 (filled circles). For the ArifloR competition experi-
ment, all cells were co-incubated with 5 AM rolipram. In the RP73401
competition, cells were also co-incubated with 5 AM rolipram. MeanF S.D.
(n= 3 separate experiments).
Fig. 8. Rolipram and RP73401 on PDE4A4–GFP expression in CHO cells. (a) Dose–response curves for the increase in total GFP in CHO/PDE4A4–GFP
cells following treatment with racemic rolipram (filled circles) or RP73401 (open circles). Cells were incubated for 16 h with the inhibitors and then read for
total GFP. There is no correction for cell number in these measurements. Values are represented as percentage increase above the lowest concentration tested,
being ‘100%’ at 0.03 AM rolipram and 0.003 AM RP73401. There are no accretion foci formed in cells at these lowest doses of inhibitor, and none are formed
at any successively higher dose of RP73401. Each value expressed as meanF S.D. (n= 4). In (b) are shown CHO cells expressing PDE4A4–GFP with images
taken at the indicated times after the re-addition of rolipram (10 AM), 3 h after washout of prior inhibitor treatments. In the first panel of images (i, ii, iii), cells
had previously been exposed to RP73401 (3 AM). No accretion foci formed with this treatment, but PDE4A4–GFP increased per cell (as shown in Fig. 8a).
They were then washed and incubated in fresh medium without rolipram for 3 h at 37 jC. In the second panel of images (iv, v, vi), cells were pretreated with
rolipram (10 AM) for 16 h to induce accretion foci. They were then washed and incubated in fresh medium without rolipram for 3 h at 37 jC so that all
rolipram-induced foci of PDE4A4–GFP had dispersed. Cells were returned to incubator conditions between images.
R. Terry et al. / Cellular Signalling 15 (2003) 955–971 965
R. Terry et al. / Cellular Signalling 15 (2003) 955–971966
model for that of all four subfamilies. The PDE4 catalytic
centre [43] features a binuclear Zn2 +–Mg2 + motif at one
end of a deep substrate-binding cleft (Fig. 9). The Zn2 +
centre is located more deeply in the cleft and is tightly
engaged by four direct ligand interactions from protein
residues. One of these residues, corresponding to Asp474
in PDE4A4, together with a hydroxide ion, acts as a
bridging ligand between the two metal centres. The Mg2 +
coordination shell is completed by solvent ligands, which
are networked through hydrogen bonds to four adjacent
residues, corresponding to His473, Glu503, His506 and
Thr544 in PDE4A4. These residues, which grip the Mg2 +
ion, are contributed by the N- and C-terminal ends of
helices-10 and -11 in the PDE4B crystal structure and by
the loop between helices-7 and -8. The sequence spanning
the 7/8-loop through helix-11 exhibits 92% residue identity
between PDE4A and PDE4B and where differences are
present, these are only conservative substitutions. The
catalytic centre thus comprises a tightly bound Zn2 + ion
coupled to a more loosely held Mg2 + centre. It has been
suggested [44,45] that conformational changes in the PDE4
catalytic unit that can be detected through changes in
binding affinity for rolipram may be effected by alterations
in liganding of the Mg2 + ion. Thus, amino acids serving to
ligand Mg2 +, either directly or indirectly through bound
H2O, may provide a potential route whereby the binding of
rolipram in the active site might ‘transmit’ conformational
changes that trigger the series of events leading to foci
generation. In this regard, His506 is of interest because it is
closely linked to the bound Mg2 + ion through a hydrogen
bond from one of its imidazole ring nitrogens to one of the
Mg2 + water ligands and because it is located on helix-10,
which is partially surface-exposed (Fig. 9a–c). Not only
this, but on the other side of the His506 imidazole ring, the
NH group H-bonds to the main chain amide carbonyl O
atom of Val475 (Fig. 9b), a residue located on the loop
between helices-7 and -8, which is also partially surface-
exposed (Fig. 9c) and which contributes the bridging metal
ligand, Asp474. This network has the potential to provide a
means for ‘inside-out’ signalling, where conformational
changes occurring in the active site could be transmitted
to the molecule surface. To investigate this possibility, we
set out to perturb the catalytic site at the position of His506
in such a way that we hoped might affect rolipram-induced
Fig. 9. The PDE4 catalytic site and ‘inside-out’ signalling elicited by rolipram bin
core catalytic unit with the Mg2 + ion held between helices-10 and -11. cAMP is ma
the model of Xu et al. [43]; the cleft is defined by the Zn2 +–Mg2 + motif and key
detail of the Mg2 + binding environment viewed from the N-terminus of helix-10; H
residues in PDE4A4 are H506 and V475. The location of the surface exposed H3
PDE4A4 are H505 and D476. (c) Two views of the PDE4 core catalytic unit with
helix-10 (red) and 7/8-loop (green) are marked; H505 on helix-10 is highlighted
behind the blue surface with access (orthogonal to the view) as indicated by the arr
the floor of a pronounced surface cleft. (d) Model for the purported interactions wit
(e) Dimerisation interface of the PDE4 core catalytic unit. The structure of the
highlighting one molecule in yellow and one in white. The white molecule has
orthogonal to helix-10. The colour coding on the white molecule is the same as th
(purple).
foci formation while retaining catalytic activity. In doing
this, we reasoned that Asn might serve as a surrogate for a
neutral His residue (Fig. 9d) and, therefore, that His506Asn
mutation might still support catalytic activity but would
subtly alter the catalytic site sufficiently to perturb a
potential rolipram-induced conformational change mediated
through the Mg2 + ion and His506 on helix-10. Here we
show that the His506Asn mutant form of PDE4A4–GFP is
clearly catalytically active, albeit this is reduced to around
half that exhibited by the wild-type enzyme (Table 1). In
addition, as we had hoped, the sensitivity of the His506Asn
mutant to activation by Mg2 + is attenuated by around eight-
fold, relative to the wild-type enzyme (Table 1). Strikingly,
however, in dramatic contrast to the wild-type enzyme, the
His506Asn mutant singularly fails to form accretion foci in
cells challenged with rolipram (5 up to 100 AM, 10–48 h;
data not shown), while still exhibiting a similar sensitivity to
wild-type PDE4A4 in its ability to be inhibited by rolipram
(Table 1).
Given that the NH group of the His506 imidazole ring H-
bonds to the main chain amide carbonyl O atom of Val475
(Fig. 9b), we set out to evaluate the potential importance of
Val475 in foci formation. In doing so, we decided to mutate
Val475 to a charged aspartate residue on the basis that the
change in polarity of the residue at this position might
sufficiently perturb the main chain to affect the hydrogen-
bonded interaction of Val475 with His506. We see here that
the Val475Asp mutant is also catalytically active, albeit with
an activity some 23F 4% (meanF S.D.; n = 6) that of the
wild-type enzyme (Table 1). Consistent with the proposed
‘coupling’ of the interaction of Val475 with the Mg2 +-
interacting His506, we noted a dramatic decrease in sensi-
tivity of the Val475Asp mutant to activation by Mg2 + of
some 11-fold (Table 1). Furthermore, as with the His506Asn
mutant, the Val475Asp mutant singularly fails to form
accretion foci in cells challenged with rolipram (5 up to
100 AM, 10–48 h; data not shown), despite exhibiting
similar sensitivity to wild-type PDE4A4 in its ability to be
inhibited by rolipram (Table 1).
We also evaluated His505, a surface-exposed residue
adjacent to His506 on helix-10 which, although not liganded
to the bound Mg2 +, forms a salt bridge to the side chain of
Asp476, which is adjacent to Val475 and thus provides
another link between helix-10 and the loop between heli-
ding. (a) Shows the crystal structure and three subdomains of the PDE4B2
nually positioned into the active site and inhibitor-binding cleft according to
residues marked (equivalent PDE4A4 numbers in parenthesis). (b) Shows
307 is hydrogen bonded to Mg2 +-bound water and V276; the corresponding
06, and its interaction with D277 is also highlighted; equivalent residues in
surface rendering taken orthogonal to axis of helix-10. The exposed regions
(blue). The catalytic cleft (obscured from view) is located approximately
ow. View (ii), as observed along the indicated ‘view’ arrow, reveals H505 in
h mutation of the histidine at position 307 (H506 in PDE4A4) to asparagine.
dimeric PDE4D core catalytic unit [53] is shown with surface rendering
a similar orientation to view (i) in image (c) and is viewed approximately
at in image (c) except that the surface of helix-9 is additionally highlighted
R. Terry et al. / Cellular Signalling 15 (2003) 955–971968
ces-7 and -8 (Fig. 9b). Consistent with the suggested impor-
tance of such a link, the His505Asnmutant form again fails to
form accretion foci in cells challenged with rolipram (5 up to
100 AM, 10–48 h; data not shown). However, His505Asn is
catalytically functional, with an activity that is around half
that of wild-type PDE4A4. Intriguingly, it is not only still
inhibited by rolipram in a manner not too dissimilar from that
of wild-type PDE4A4 (Table 1), but also shows but little
difference in its sensitivity to activation by Mg2 + (Table 1).
This suggests that ablation of rolipram-induced foci genera-
tion is not directly related to attenuated sensitivity to Mg2 +.
Specificity for PDE4A4 in foci formation by rolipram
indicates a role for the unique N-terminal region of this
isoform in the response. It is possible that ‘inside-out’
signalling achieved by the rolipram-induced conformational
change in the catalytic unit exerts itself by altering an
interaction of the catalytic unit with the N-terminal region
of PDE4A4. We sought to provide an initial evaluation of
this by trying to disrupt the interaction between the N-
terminal portion of PDE4A4 with the catalytic unit. To do
this, we focussed on LR2, the 25 amino acid region that
connects UCR2, and through it both UCR1 and the unique
N-terminal of PDE4A4, to the catalytic unit. We have
previously shown [25] that the Db-LR2-PDE4A4 construct,
which has a small eight amino acid (APRPRPSQ) deletion
in LR2, is catalytically active. However, Db-LR2-PDE4A4
causes a conformational change in PDE4A4 that ablates the
ability of LYN SH3 interaction to alter the kinetics of
rolipram inhibition of this isoform [25], making examina-
tion of this mutant particularly pertinent. We show here that
Db-LR2-PDE4A4–GFP exhibits an attenuated ability to
form foci in response to challenge with rolipram (Fig. 10).
That it is not ablated is consistent with our demonstration
that PDE4A5, the rat homologue of PDE4A4 can form foci
(Fig. 3d) but does not possess the proline and arginine-rich
region deleted in the Db-LR2-PDE4A4–GFP construct
[25]. This suggests that the appropriate orientation of the
Fig. 10. Attenuated foci formation seen upon disruption of LR2.
Experiments were done as described in Fig. 6a. Each value is the
meanF S.D. (n= 6). Data are shown for wild-type PDE4A4-chimera (filled
circles) and the Db-LR2-PDE4A4–GFP chimera (open triangles).
unique N-terminal portion of PDE4A4 to the catalytic unit is
required for efficient foci formation to ensue.
4. Discussion
Here, for the first time, we show that the PDE4 selective
inhibitor, rolipram causes the profound redistribution of the
PDE4A4 isoform in a dose-dependent and reversible fash-
ion. To facilitate the analysis of this effect and to follow it in
living cells, we have exploited a PDE4A4 chimera formed
with GFP. Indeed, GFP has proven itself as a powerful tool
to shed insight into the workings of many biological
systems as it allows an effective means of identifying
dynamic events occurring in living cells [23]. Of course,
as with any tool used in investigative biology, it is not
without its potential problems. We appreciate that the data
presented in this study may well appear to be provocative as
the effects reported are simply so very striking. Interesting-
ly, the formation of similar, striking cytoplasmic foci has
recently been reported in the identification of cytoplasmic
processing bodies, which among other things accumulate
translationally repressed mRNAs, as evaluated in studies
using both GFP chimera and native protein [46–48]. We
would contend that the foci formation detected here reflects
a specific event that is associated with and triggered by the
PDE4A4 component rather than some simple artefact asso-
ciated with either GFP or the various other protein tags used
to generate chimeras with PDE4A4. Our reasons for think-
ing this are that (i) foci formation can simply be reversed by
removal of rolipram, with subsequent rolipram re-addition
causing re-formation of foci; (ii) rolipram does not cause the
redistribution of GFP itself; (iii) rolipram induces foci
formation in PDE4A4 tagged with various other proteins
and with wild-type PDE4A4; (iv) various single point
mutations in the PDE4A4 catalytic unit specifically ablate
foci formation while still allowing rolipram to inhibit
PDE4A4 catalytic activity; (v) foci formation shows spec-
ificity for PDE4A4 and its homologue, PDE4A5, rather than
all PDE4 species; (vi) certain PDE4 inhibitors, exemplified
by Ariflo and RP73401, not only are unable to cause
accretion foci to form, but will both disperse rolipram pre-
formed foci and compete out the foci-forming action of
rolipram in a dose-dependent fashion; (vii) elevation of
intracellular cAMP levels achieved through forskolin-stim-
ulation of adenylyl cyclase competes out foci formation,
presumably by displacing rolipram from the active site of
PDE4A4 and (viii) foci formation shows discrimination
between the two enantiomeric forms of rolipram, with (R)-
(�)-rolipram preferred over (S)-(+)-rolipram. Thus, not only
is PDE4A4 essential for foci formation but rolipram is
needed in the active site of PDE4A4 for it to occur (Fig. 11).
Various investigators [3,36,37,41,44] have suggested that
whereas rolipram, Ro20–1724 and RS25344 appear able to
‘sense’ so-called high affinity (HARBS) and low affinity
(LARBS) PDE4 conformers, this is not seemingly true for
Fig. 11. Schematic for the generation and dispersal of accretion foci of PDE4A4–GFP in cells. A model summarising key aspects of the changes that lead up to
rolipram-induced foci generation in PDE4A4. Binding of rolipram to the catalytic unit is envisaged to cause a conformational change that is transmitted to the
surface of the molecule by ‘inside-out’ signalling. This change in conformation, together with the unique N-terminal region of PDE4A4 promotes the
aggregation of this isoform. Such a process requires active protein synthesis, which may be to provide either directly or indirectly interacting protein(s) needed
for foci formation. We note that at low concentrations of PDE4A4, this process may be curtailed at the level of either dimerisation or lower value oligomers.
Mass action would then indicate that the formation of larger foci would be promoted at higher expression levels of PDE4A4 or in localised regions of cells
where PDE4A4 expression is compartmentalised.
R. Terry et al. / Cellular Signalling 15 (2003) 955–971 969
ArifloR and RP73401, which have been suggested primarily
interact with low affinity conformers [40,41]. Our data
extend this notion by suggesting that inhibitors such as
rolipram, Ro20–1724 and RS25344, on binding within the
catalytic site of PDE4A4, are themselves able to stabilise a
very particular conformational state of the catalytic unit of
this isoform. That various other PDE4 selective inhibitors,
as exemplified here by ArifloR and RP73401, cannot do
this suggests that a specific region of the active site is
involved. Here we detect this by the accretion of PDE4A4
chimeras into foci.
Specificity of this effect for PDE4A4 [30] and its rodent
homologue, PDE4A5 [49] implies that the N-terminal
region, which is unique to these isoforms, is intimately
involved in foci formation along with the PDE4 catalytic
unit, where rolipram binds in order to trigger this response.
That not all PDE4 selective inhibitors elicit such foci
formation suggests that the particular mode of binding of
certain of these inhibitors to the catalytic unit causes a
distinct conformational change. Indeed our discrete set of
mutations within the catalytic unit shows that it is possible
to ablate the ability of rolipram to form foci while retaining
its ability to inhibit catalytic activity. Such observations are
consistent with the notion that enzyme inhibition and foci
formation are not directly connected; as is evident from the
identification of non-foci-forming inhibitors. A consequence
of this proposal (Fig. 11) is that by simply competing out
rolipram from the catalytic site, then both elevated cAMP
levels and non-foci-forming PDE4 inhibitors would be able
to antagonise rolipram-induced foci formation and to dis-
perse rolipram pre-formed foci, as indeed is shown here.
Given the important role of PDE4 in cellular signalling
pathways and the enzyme’s potential therapeutic signifi-
cance, we would hope that the novel distinction between
PDE4 inhibitors, revealed by their ability to induce foci
formation, and the proposed conformational differences in
the enzyme associated with their binding might provoke
structural studies of PDE4A4 done in the presence of
compounds that either do or do not cause foci formation.
Together, with the indicated involvement of the enzyme’s
N-terminus, we consider that the binding of particular
inhibitors, such as rolipram, to the active site might induce
specific conformational changes by movement in the metal
centres, which is relayed through the ligand residues and
propagated to the exterior of the core catalytic unit. We call
this action, where rolipram binding to the catalytic site
effects a signalling action transmitted to the molecule
surface, ‘inside-out’ signalling. In this regard, other than
the catalytic site itself, the surfaces of the 7/8-loop, N-
terminus of helix-10 and C-terminus of helix-11 constitute
the most proximal externally exposed regions of the enzy-
me’s core catalytic unit to the Mg2 + centre (Fig. 9c), and
provide the most direct route for transmitting, to the
molecule surface, conformational changes consequent upon
rolipram binding at the active site. Although clearly pro-
scribed, our single point mutation studies, which identify
residues whose mutation ablates foci formation while still
allowing rolipram to cause inhibition of catalytic activity,
R. Terry et al. / Cellular Signalling 15 (2003) 955–971970
hint at the possible importance of this route. These identify
His506, which is hydrogen bonded through one of its
imidazole ring nitrogen atoms to a water molecule ligand
on the Mg2 + ion in the catalytic site. Also His505, which
although surface-exposed on helix-10, interacts with the
loop between helices-7 and -8, as does His506 (Fig. 9c).
This loop, like helix-10, contributes a Mg2 +-binding residue
(Asp474) and forms part of the exposed surface of the PDE4
catalytic site (Fig. 9b). Consistent with the suggested
importance of the 7/8-loop in relaying the presence of bound
rolipram to the protein surface, the Val475Asp mutant,
while again catalytically active and inhibited similarly to
wild-type PDE4A4 by rolipram, failed to form foci in
rolipram-challenged cells. This may be due to the disruption
of interaction with His506, and thereby subtle interference
in the positioning and/or movement of helix-10 and the 7/8-
loop, although without precise structural data, we cannot
exclude a more general disruption of the catalytic unit by
this mutation. However, that this mutant PDE4A4 is both
active and sensitive to inhibition by rolipram might indicate
that a subtle perturbation has been achieved. It is intriguing
to note that His505 is found exposed at the base of a dramatic
groove that runs across the surface of the 7/8-loop and
around the circumference of helix-10 (Fig. 9c), which might
provide a docking site for an interacting protein. Alterna-
tively, it might accommodate either the PDE4A4 unique N-
terminal region or the regulatory UCR1/2 modules, thereby
providing a link between the catalytic unit and the unique N-
terminal region of PDE4A4. Certainly, the unique N-termi-
nal portion of PDE4A4 appears to be essential for foci
generation since other PDE4 isoforms (PDE4B2, PDE4C2,
PDE4D3 and PDE4D5) fail to form accretion foci upon
rolipram challenge. Moreover, perturbation of its linkage to
the catalytic unit by deletion of a small portion within the
region between UCR2 and the catalytic unit (LR2; [25]), as
seen in the Db-LR2-PDE4A4–GFP chimera, clearly makes
foci formation somewhat less efficient. Our investigations on
this deletion construct also indicate that ligation of an SH3-
domain-containing protein to the proline and arginine-rich
region LR2 of PDE4A4 [25] is not required for rolipram-
induced foci formation to ensue.
The molecular basis of this redistribution process, and
whether this action of rolipram exploits a natural regulatory
process triggered either by the binding of an as yet uniden-
tified ligand or a post-translational modification remains to
be ascertained. However, while it is clear that foci genera-
tion is not driven by either PKA action or increased cAMP
levels, active protein synthesis is required. Indeed when
cells are washed free of rolipram to disperse foci, their
rechallenge with rolipram causes foci to reform more
rapidly than in naı̈ve cells (Fig. 5). This implies that the
requirement for protein synthesis is to induce a protein(s)
essential for foci formation and this is clearly specific for
rolipram since pretreatment with non-foci-forming PDE4
inhibitors does not impart such memory (Fig. 5). Thus,
rolipram-elicited ‘inside-out’ signalling might trigger the
induction of specific proteins required for foci formation.
However, defining any such action poses a considerable
technical challenge that lies beyond the scope of this study.
Unfortunately, in the various cultured cells, we have
examined so far, endogenously expressed PDE4A4 is found
at such low protein levels that it militates against its analysis
in immunofluorescence studies. Thus, either technological
advances or the identification of a suitable cell type are
required in order to be able to evaluate whether endoge-
nously expressed PDE4A4 undergoes rolipram-mediated
relocalisation to foci. It is conceivable that at the low levels
of native expression of PDE4A4 and in the absence of
additional protein tags the reversible aggregation may be
less acute, even perhaps restricted to a facilitation of
dimerisation. Dimerisation has been suggested to occur for
certain PDE4 isoforms [50–52] and, indeed, a dimerisation
interface has been identified in a recent X-ray crystal
structure of the PDE4D core catalytic unit complexed with
the inhibitor, zardaverine [53]. The dimer interface region
identified in this study is highly conserved across all PDE4
isoforms and involves residues on the surface-exposed
regions of helices-9 and -10, which are close to both
His505 and the region that we have identified in PDE4A4
in connection with the ‘inside-out’ signalling of rolipram
(Fig. 9e). An important question, then, is whether rolipram
and related inhibitors that promote foci formation stabilise
the dimeric interface and whether, following the induction
of protein synthesis, dimeric PDE4A4 units further aggre-
gate to form foci (Fig. 11). Clearly, in this study, cells have
been designed to overexpress PDE4A4 in order to detect
and analyse it, thus any change in intracellular distribution
of the native enzyme may then be far subtler. However, if
rolipram-induced relocalisation of PDE4A4 should occur
natively, then it may have functional consequences that
extend further than simply the spatial re-organisation of
cAMP degradation caused, for example, through any ac-
companying translocation of PDE4A4-interacting proteins.
Thus, particular PDE4 selective inhibitors, such as rolipram,
may exert effects on cells through PDE4A4 that extend
beyond simple catalytic inhibition and any consequent
elevation of cAMP levels (Fig. 11). Such cAMP-indepen-
dent effects exerted by this ‘inside-out’ signalling may not
only take the form of the induction of specific proteins but
also extend to additional functional consequences caused by
the cellular relocalisation of any PDE4A4-associated pro-
teins to foci. Even without unequivoval confirmation of this
possibility, it is clear that specific inhibitor-dependent con-
formational changes in the structure of PDE4A4 must be
occurring. The technology described here offers a novel and
sensitive means of sub-categorising PDE4 inhibitors and
probing for conformational changes in PDE4A4 induced by
inhibitor binding. In this regard, we have pinpointed a core
group of networked residues on helix-10 and the 7/8-loop of
the core catalytic unit that may provide part of a conforma-
tional relay, communicating the presence of a bound inhib-
itor from the catalytic site and its bound Mg2 + centre to the
R. Terry et al. / Cellular Signalling 15 (2003) 955–971 971
protein surface. Whether the functional outcome of the
induced conformational changes in PDE4A4-associated foci
formation relates to the side effects seen with various PDE4
inhibitors and which have compromised their therapeutic
deployment or whether it provides added therapeutic op-
portunity remains to be seen. Of course foci formation may
reflect the activation of a normal process involved in
regulating PDE4A4 that happens, as we show here, also
to be activated by rolipram. We trust that the analyses
presented here will serve to stimulate studies aimed at
addressing these various questions.
Acknowledgements
M.D.H. is funded by the MRC (UK) G8604010 and the
European Union (QLG2-CT-2001-02278; QLK3-CT-2002-
02149). BT thanks the EC for support in grant QLG2-CT-
2001-02278. Our special thanks go to Dr. Sara P. Bjoern and
her team in BioImage for expert advice on, and preparation
of, PDE4A4B chimeras and mutants used in this study, and
also to Grith Hagel for developing the extraction procedure
used to measure immobile PDE4A4–GFP in cells.
References
[1] Beavo JA, Brunton LL. Nat Rev, Mol Cell Biol 2002;3(9):710–8.
[2] Conti M, Jin SLC. Prog Nucleic Acid Res 1999;63:1–38.
[3] Houslay MD. Prog Nucleic Acid Res Mol Biol 2001;69:249–315.
[4] Manganiello VC, Degerman E. Thromb Haemost 1999;82(2):
407–11.
[5] Francis S, Turko I, Corbin J. Prog Nucleic Acid Res Mol Biol 2001;
65:1–52.
[6] Colledge M, Scott JD. Trends Cell Biol 1999;9(6):216–21.
[7] Houslay MD, Milligan G. Trends Biochem Sci 1997;22(6):217–24.
[8] Zhang J, Ma Y, Taylor SS, Tsien RY. Proc Natl Acad Sci U S A 2001;
98(26):14997–5002.
[9] Zaccolo M, Pozzan T. Science 2002;295(5560):1711–5.
[10] Jurevicius J, Fischmeister R. Proc Natl Acad Sci U S A 1996;93(1):
295–9.
[11] Barnette MS, Underwood DC. Curr Opin Pulm Med 2000;6(2):
164–9.
[12] Giembycz MA. Monaldi Arch Chest Dis 2002;57(1):48–64.
[13] Torphy TJ. Am J Respir Crit Care Med 1998;157(2):351–70.
[14] Huang Z, Ducharme Y, Macdonald D, Robichaud A. Curr Opin Chem
Biol 2001;5(4):432–8.
[15] Burnouf C, Pruniaux MP. Curr Pharm Des 2002;8(14):1255–96.
[16] Doherty AM. Curr Opin Chem Biol 1999;3(4):466–73.
[17] Houslay MD, Adams DR. Biochem J 2003;370:1–18.
[18] Shakur Y, Pryde JG, Houslay MD. Biochem J 1993;292(3):677–86.
[19] Yarwood SJ, Steele MR, Scotland G, Houslay MD, Bolger GB. J Biol
Chem 1999;274(21):14909–17.
[20] Tasken KA, Collas P, Kemmner WA, Witczak O, Conti M, Tasken K.
J Biol Chem 2001;276(25):21999–2002.
[21] Dodge KL, Khouangsathiene S, Kapiloff MS, Mouton R, Hill EV,
Houslay MD, et al. EMBO J 2001;20(8):1921–30.
[22] Baillie GS, Huston E, Scotland G, Hodgkin M, Gall I, Peden AH, et al.
J Biol Chem 2002;277(31):28298–309.
[23] Tsien RY. Ann Rev Biochem 1998;67:509–44.
[24] MacKenzie SJ, Baillie GS, McPhee I, MacKenzie C, Seamons R,
McSorley T, et al. Br J Pharmacol 2002;136(3):421–33.
[25] McPhee I, Yarwood SJ, Huston E, Scotland G, Beard MB, Ross AH,
et al. J Biol Chem 1999;274:11796–810.
[26] Huston E, Pooley L, Julien J, Scotland G, McPhee I, Sullivan M, et al.
J Biol Chem 1996;271:31334–44.
[27] Thompson WJ, Appleman MM. Biochemistry 1971;10:311–6.
[28] Laemmli UK. Nature (Lond) 1970;222:680–2.
[29] Bradford M. Anal Biochem 1976;72:248–54.
[30] Bolger G, Michaeli T, Martins T, St John T, Steiner B, Rodgers L,
et al. Mol Cell Biol 1993;13(10):6558–71.
[31] Livi GP, Kmetz P, McHale MM, Cieslinski LB, Sathe GM, Taylor DP,
et al. Mol Cell Biol 1990;10(6):2678–86.
[32] Engels P, Fichtel K, Lubbert H. FEBS Lett 1994;350(2–3):291–5.
[33] Seybold J, Newton R, Wright L, Finney PA, Suttorp N, Barnes PJ,
et al. J Biol Chem 1998;273(32):20575–88.
[34] Schwartz R, Helmich B, Spector D. J Virol 1996;70:6955–66.
[35] Hunninghake G, Monick M, Liu B, Stinski M. J Virol 1989;63:
3026–33.
[36] Alvarez R, Sette C, Yang D, Eglen RM, Wilhelm R, Shelton ER, et al.
Mol Pharmacol 1995;48(4):616–22.
[37] Souness JE, Rao S. Cell Signal 1997;9:227–36.
[38] Houslay MD, Sullivan M, Bolger GB. Adv Pharmacol 1998;44:
225–342.
[39] Torphy TJ, Barnette MS, Underwood DC, Griswold DE, Christensen
SB, Murdoch RD, et al. Pulm Pharmacol Ther 1999;12(2):131–5.
[40] Barnette MS, Christensen SB, Essayan DM, Grous M, Prabhakar U,
Rush JA, et al. J Pharmacol Exp Ther 1998;284(1):420–6.
[41] Jacobitz S, McLaughlin MM, Livi GP, Burman M, Torphy TJ. Mol
Pharmacol 1996;50:891–9.
[42] Barnette MS, Christensen SB, Underwood DC, Torphy TJ. Pharmacol
Commun 1996;8(1):65–73.
[43] Xu RX, Hassell AM, Vanderwall D, Lambert MH, Holmes WD,
Luther MA, et al. Science 2000;288(5472):1822–5.
[44] Laliberte F, Han Y, Govindarajan A, Giroux A, Liu S, Bobechko B,
et al. Biochemistry 2000;39(21):6449–58.
[45] Liu S, Laliberte F, Bobechko B, Bartlett A, Lario P, Gorseth E, et al.
Biochemistry 2001;40(34):10179–86.
[46] Anderson P, Kedersha N. J Cell Sci 2002;115:3227–34.
[47] Kedersha N, Chen S, Gilks N, Li W, Miller IJ, Stahl J, et al. Mol Biol
Cell 2002;13:195–210.
[48] Sheth U, Parker R. Science 2003;300:805–8.
[49] McPhee I, Pooley L, Lobban M, Bolger G, Houslay MD. Biochem J
1995;310(3):965–74.
[50] Rocque WJ, Tian G, Wiseman JS, Holmes WD, Zajac-Thompson I,
Willard DH, et al. Biochemistry 1997;36(46):14250–61.
[51] Rocque W, Holmes W, Patel I, Dougherty R, Ittoop O, Overton L,
et al. Protein Expr Purif 1997;9:191–202.
[52] Richter W, Conti M. J Biol Chem 2002;12:12.
[53] Lee ME, Markowitz J, Lee J-O, Lee H. FEBS Lett 2002;530:53–8.