From the Section for Molecular Signaling, Department
of Cell and Molecular Biology, Lund University, S-221 00 Lund, Sweden
and ¶ Pulmonary/Critical Care Medicine Branch, NHLBI, National
Institutes of Health, Bethesda, Maryland 20892
cAMP and cGMP mediate biological responses
initiated by diverse extracellular signals. By catalyzing hydrolysis of
the 3
Mammalian PDEs share a common structural organization, with a conserved
catalytic core (~270 amino acids) usually located in the C-terminal
half (Fig. 1) (4). This region is much more similar within an
individual PDE family (>80% amino acid identity) than between
different PDE families (~25-40% identity) (1-4). The catalytic
core is thought to contain common structural elements important for
hydrolysis of the cyclic nucleotide phosphodiester bond, as well as
family-specific determinants responsible for differences in substrate
affinities and inhibitor sensitivities among the different gene
families. It contains a PDE-specific sequence motif,
HD(X)2H(X4)N, and two
consensus Zn2+-binding domains, the second of which
overlaps the PDE motif (3, 5). PDE5 contains tightly bound
Zn2+, which supports catalytic activity (5). The precise
role of Zn2+ or other divalent cations in catalytic
function of other PDEs has not been defined. Mutagenesis of the first
histidine of the PDE sequence motif abolished activity of a recombinant
PDE4 expressed in Escherichia coli (6). Histidine- and
sulfhydryl-modifying reagents inhibited PDE3 activity (7).
The widely divergent N-terminal portions of PDEs (Fig. 1) contain
determinants that confer regulatory properties specific to the
different gene families, e.g. calmodulin-binding domains (PDE1); two non-catalytic cyclic nucleotide-binding domains (PDEs 2, 5, and 6); N-terminal membrane-targeting (PDE4) or hydrophobic membrane-association (PDE3) domains; and calmodulin (PDE1)-, cyclic AMP
(PDEs 1, 3, and 4)-, and cGMP (PDE5)-dependent protein
kinase phosphorylation sites, etc. (Fig. 1) (1-3). Most cells contain representatives of several PDE families in different amounts, proportions, and subcellular locations (1-3). In some instances a
specific PDE regulates a unique cellular function, e.g.
photoreceptor PDE6 in cGMP-dependent initiation of visual
transduction. In individual cells, different PDEs, with their different
responses to regulatory signals, participate in integrating multiple
inputs in the complex modulation and termination of cyclic nucleotide
signals and responses, e.g. their magnitude and duration,
their functional and spatial compartmentation, and their attenuation by
short-term feedback or long-term desensitization.
Distinctive Characteristics of PDE3s PDE3s, purified to apparent homogeneity from a variety of
tissues, can be distinguished from other PDEs by their high affinities for both cAMP and cGMP, with Km values in the range
of 0.1-0.8 µM and Vmax for cAMP
4-10 times higher than that for cGMP (1-3, 8).
PDE3 is for several reasons often referred to as the cGMP-inhibited
PDE. When different PDEs were first identified, two types (now
classified as PDE3 and PDE4 (3)) that exhibited a high affinity for
cAMP were isolated from various tissues. PDE3, but not PDE4, exhibited
a high affinity for both cAMP and cGMP. As might be predicted from
their Km values, cAMP and cGMP were mutually
competitive substrates for PDE3. In contrast, cGMP was hydrolyzed
poorly by, and did not inhibit, PDE4. Thus, PDE3 was called the
cGMP-inhibited PDE to distinguish it from PDE4. In fact, some
biological effects of endogenous cGMP may be mediated by inhibition of
PDE3, which results in increased cAMP and activation of
cAMP-dependent protein kinase (protein kinase A) (1-3).
For example, in rabbit platelets (9), mouse thymocytes (10), and human
atrial and frog ventricular myocytes (11), nitrovasodilators (which
release nitric oxide and activate guanylyl cyclase) increase cAMP, at
least in part, by increasing cGMP, which inhibits PDE3. PDE2 isoforms,
which are allosterically activated by cGMP, also serve as a locus for
"cross-talk" between cAMP and cGMP signaling systems. PDE2 is
highly concentrated in bovine adrenal glomerulosa cells where atrial
natriuretic factor inhibits cAMP-stimulated aldosterone biosynthesis,
at least in part, by stimulating guanylyl cyclase and increasing cGMP,
which activates PDE2 leading to a decrease in cAMP and protein kinase A
activity (12).
PDE3 isoforms are also characterized by their sensitivity to several
specific inhibitors and drugs, including cilostamide, enoximone, and
lixazinone (reviewed in Refs. 13-15), compounds relatively selective
for PDE3, with Ki and IC50 values at
least 10-100-fold lower for PDE3 than for other PDE families. The
availability of family-specific PDE inhibitors, especially for PDEs 3, 4, and 5 (Fig. 1), has facilitated understanding of functions of
individual PDEs in regulating specific cyclic nucleotide-mediated processes, e.g. PDE3s in regulation of certain
cAMP-modulated processes, including stimulation of myocardial
contractility, inhibition of platelet aggregation, relaxation of
vascular and airway smooth muscle, and inhibition of proliferation of
T-lymphocytes and cultured vascular smooth muscle cells (13-17). The
pharmaceutical industry has exhibited considerable interest in
developing specific inhibitors of individual PDE families and
subfamilies as therapeutic agents to replace widely used but
nonselective PDE inhibitors such as theophylline. That subject is,
however, beyond the scope of this review.
A third important general characteristic of PDE3s (discussed in more
detail below) involves their phosphorylation and short-term activation
in response to insulin as well as to agents that increase cAMP in
adipocytes, hepatocytes, and platelets (reviewed in Ref. 18). Other
PDEs, in addition to PDE3, are also regulated by phosphorylation
(1-3). For example, phosphorylation of PDE1A by
Ca2+/calmodulin-dependent protein kinase or
PDE1B by protein kinase A reduces affinity of both PDE1 isoforms for
calmodulin. Binding of cGMP to non-catalytic binding sites in the
regulatory domain of PDE5 enhances phosphorylation of PDE5 by
cGMP-dependent protein kinase (protein kinase G).
Phosphorylation of Tissue-specific Expression, Structure, and Subcellular Location of
PDE3 Isoforms cDNAs for two PDE3 isoforms (recently classified as PDE3A and
PDE3B, respectively (3)) have been cloned from human (H) and rat (R)
libraries (19, 20). Deduced sequences of PDE3A (or -B) from different
species (rat and human) are more similar than are those of PDE3A and -B
from the same species (Fig. 2). The human isoforms,
HPDE3A and -B, are products of different genes on chromosomes 12 and
11, respectively. Two HPDE3A mRNA species of ~7.6 and ~4.4
kilobases, encoding predicted ~125- and 80-kDa proteins, may be
transcribed in a tissue-specific manner, using different initiation
sites in the HPDE3A gene (21).
In situ hybridization and Northern blot hybridizations
demonstrated overlapping but distinct tissue and cellular distributions of (rat) RPDE3A and RPDE3B mRNAs (20, 22). RPDE3B mRNA was prominent in white and brown adipose cells, hepatocytes, renal collecting duct epithelium, and developing spermatocytes; RPDE3A mRNA was more abundant in heart and vascular smooth muscle (22). The distribution of RPDE3A mRNA in developing rat brain was
heterogenous, whereas RPDE3B mRNA was uniformly present in germinal
neuroepithelium and mature neurons (23). PDE3B (not PDE3A) mRNA and
enzyme activity (associated with adipocyte particulate fractions) was
found in cultured murine 3T3-L1 adipocytes but not undifferentiated
3T3-L1 fibroblasts (24). Human platelet PDE3 is a PDE3A isoform (25). These and other findings suggest that PDE3A and -B likely exhibit cell-specific differences in properties and regulation and may serve
cell-specific functions.
Cell-specific expression of different members of the PDE1 family has
also been reported (3). In situ hybridization demonstrated that of three different PDE1s in mouse and rat brain, PDE1A mRNA is
abundant in cortex and portions of the hippocampus, PDE1B mRNA in
the striatal region and dentate gyrus, and PDE1C in olfactory neuronal
epithelia (3). Selective expression of different representatives of the
same PDE family (or of different families) in different and limited
populations of cells (1-3) has important implications not only for
regulation of cyclic nucleotide concentrations and their biological
effects in specific cells but also in targeting of specific PDEs for
therapeutic intervention.
The structural organization of PDE3A and -3B proteins is identical. The
catalytic domain conserved among all PDEs is in the C-terminal half of
the PDE3 molecules (Fig. 1) and is followed by a hydrophilic C-terminal
region (Fig. 2). Although the catalytic domains of PDE3A and -3B are
very similar, an insert of 44 amino acids in the PDE3-conserved domain,
which does not align with sequences in the cognate domains of other PDE
families (Fig. 1), differs in PDE3A and -B isoforms (Fig. 2). This
insertion, which distinguishes PDE3 catalytic domains from those of
other PDEs and may identify subfamilies within the PDE3 family,
interrupts the first (of two) putative Zn2+-binding domains
present in the catalytic domains of all PDEs (5, 19, 20). Whether the
44-amino acid insert in PDE3 is involved in its interaction with
substrates and inhibitors remains to be established. The N-terminal
portions of PDE3A and -3B are quite divergent (Fig. 2). The N terminus
of RPDE3B is enriched in proline residues (20). Both PDE3A and -B
contain hydrophobic putative membrane-association domains with several
predicted helical transmembrane segments and several downstream
consensus sequences (RRXS) for phosphorylation by protein
kinase A (Fig. 2). In intact rat adipocytes, serine 302 in a microsomal
PDE3B is phosphorylated in response to insulin and agents that increase
cAMP (26).
A comparison of the properties of full-length and truncated PDE3A
and -B recombinants and purified platelet PDE3A after limited proteolysis indicates that the PDE3 catalytic core includes the conserved PDE domain plus some additional N- and C-terminal sequences (Fig. 2), that the N-terminal portions of PDE3A and PDE3B isoforms are
not required for PDE3 catalytic activity or sensitivity to specific
PDE3 inhibitors (Fig. 2), and that HPDE3A may be more sensitive to
inhibition by cGMP than RPDE3B (19, 25, 27, 28). Multiple
structural determinants in cGMP interact with PDE3, since
different modifications of the guanine ring altered the
IC50 values of a series of cGMP analogs for inhibition of cAMP hydrolysis (29).
PDE3s are found in both particulate and cytosolic fractions of cells
(8). The open reading frames of PDE3A and -B cDNAs predict proteins
of ~122-125 kDa (Fig. 2), consistent with monomeric sizes of
130-135 kDa for PDE3 isoforms identified in adipocyte microsomal and
cardiac sarcoplasmic reticulum fractions (30, 31). PDE3 isoforms of
~110 kDa and less have been purified from cytosolic fractions of
human platelets, bovine ventricular myocardium, and bovine aortic
smooth muscle (25, 32-34). The N-terminal hydrophobic regions, which
contain several predicted transmembrane segments, are likely to be
important in membrane association of PDE3s, since PDE3B and PDE3A
recombinants (both full-length and truncated forms, which contain
portions of the hydrophobic region) were found predominantly in
particulate fractions of Sf9 cells, whereas recombinants containing the
C-terminal catalytic core but lacking the hydrophobic sequences were
predominantly cytosolic (21, 27). Whether cytosolic PDE3 isoforms are
generated by proteolytic removal of the membrane-association region or
result from the use of alternative transcription initiation sites or
alternative mRNA splicing is unknown (21, 25, 27, 31-34).
Regulation of Adipocyte PDE3B by Insulin and Agents That
Increase cAMP In adipose tissue, activation of a microsomal PDE3B is a major
mechanism by which insulin antagonizes catecholamine-induced release of
free fatty acids, a quantitatively important energy source in mammals
(Fig. 3) (18, 35-37). Specific PDE3 inhibitors such as
cilostamide, OPC-3911, and CI-930 blocked the antilipolytic action
of insulin (38-40). Of a series of cAMP analogs, all of which
activated protein kinase A and stimulated lipolysis, insulin effectively inhibited the lipolytic effects of only those analogs that
were substrates of adipocyte PDE (41). These latter experiments suggested that activation of PDE, rather than inhibition of adenylyl cyclase or activation of protein phosphatase, was central to the antilipolytic action of insulin (41). The reduction in cAMP/protein kinase A activity that accompanies PDE3B activation results in net
dephosphorylation and decreased activity of hormone-sensitive lipase (HSL), leading to decreased hydrolysis of stored
triglyceride (Fig. 3) (18, 35-37).
Activation of the rat adipocyte PDE3B by insulin is associated
with phosphorylation of Ser-302 (based on the deduced amino acid
sequence of PDE3B (20)), catalyzed by an insulin-stimulated protein
serine kinase (PDE3IK) (Fig. 3) (26). The phosphatidylinositol 3-kinase
(PI3-K) inhibitor wortmannin blocked insulin-induced activation of
PDE3IK and phosphorylation/activation of PDE3B, as well as the
antilipolytic action of insulin (42). These results suggested that
insulin's antilipolytic signal chain, including downstream components
such as PDE3B, is regulated via receptor-mediated activation of PI3-K
(Fig. 3). As also depicted in Fig. 3, in rat adipocytes, lipolytic
hormones and agents that increase cAMP, including isoproterenol, induce
rapid phosphorylation of Ser-302 and activation of PDE3B, presumably
catalyzed by protein kinase A (26, 30, 43). Isoproterenol-induced
phosphorylation and activation of PDE3B occur over the same
concentration range as that required to activate adenylyl cyclase,
protein kinase A, and HSL (26, 30, 37, 43). This coordinate regulation
of cAMP synthesis and destruction may be important in determining steady state concentrations of cAMP and active protein kinase A (Fig.
3).
In the presence of insulin and lipolytic hormones, conditions in which
insulin reduces protein kinase A activity and inhibits lipolysis,
effects of the two agonists on phosphorylation of Ser-302 and
activation of PDE3B are more than additive (30, 37, 43). Thus, in
PDE3B, a single serine located in a protein kinase A consensus sequence
(-MFRRPS302LPCISREQ-) is phosphorylated in response to
insulin, isoproterenol, or the combination of both (26) (Fig. 3),
suggesting that the antilipolytic action of insulin involves cross-talk
between insulin- and cAMP-signaling pathways upstream of PDE3B and that
protein kinase A is involved in sensitization of the insulin-signaling pathway and activation of PDE3B (Fig. 3). cAMP-enhancing agents have
been reported to modulate responses to insulin and other growth factors
(44-46). In adipocytes, insulin-induced tyrosine phosphorylation of
insulin receptor substrate-1 (IRS-1) results in translocation of PI3-K
to IRS-1 and increased PI3-K activity (Fig. 3) (44). In this scheme,
IRS-1, PI3-K, and PDE3IK itself, as well as other unidentified
components of the signaling chain between PI3-K and PDE3IK, could be
converging points in the insulin and cAMP pathways (Fig. 3).
Mitogen-activated protein kinases and pp70S6 kinases are apparently not
involved in this cross-talk or in the activation of PDE3B by
insulin.2
As depicted in Fig. 2B, Ser-302 is just C-terminal to the
putative membrane-association domain of PDE3B. This model is consistent with earlier reports of purification of a truncated ~70-kDa PDE3 that
was released from rat hepatic membrane fractions by limited proteolysis
(48). Phosphorylation of Ser-302 might relieve inhibitory constraints
or alter conformation of the catalytic domain, since limited
proteolysis of rat adipocyte microsomal fractions from control cells
increased PDE activity to the same level as that in microsomal
fractions from insulin-treated cells (49). Ser-302 is
phosphorylated to a similar extent in intact adipocytes incubated with
hormones and in solubilized PDE3B preparations incubated with ATP and
protein kinase A (26, 47). The major site phosphorylated in
vitro, however, is Ser-427 (47), which is not phosphorylated in
intact cells (26). Whether this reflects the conformation of
membrane-bound PDE3B in intact cells or association of PDE3B with
different regulatory factors in intact and broken cells is not
known.
One could speculate that PDE3B is located in microsomal membranes in
close proximity to a docking protein for an adipocyte protein kinase
A-anchoring protein (50), involved in translocation of protein kinase A
(and other signaling molecules) to spatially segregated substrates such
as membrane-associated PDE3B. In 3T3-L1 adipocytes such spatial
segregation of PDE3B and the lipolytic machinery, resulting in PDE3B
regulation of a "pool" of cAMP related specifically to modulation
of protein kinase A and HSL, could explain the observed inhibition of
the antilipolytic effect of insulin by PDE3 inhibitors but not by PDE4
inhibitors (51). In general terms, PDEs may play an important role in
the functional or spatial compartmentation of cyclic nucleotide
signaling processes. In recent studies on regulation of
Ca2+ channels in single isolated frog ventricular myocytes,
Jurevicus and Fischmeister (52) presented evidence to suggest that
Insulin-mediated activation of PDE3 may be an important component
in insulin regulation of other cAMP-modulated processes, including
growth and differentiation and carbohydrate metabolism. In the frog
(Xenopus laevis) oocyte, stimulation of oocyte meiotic maturation by insulin, insulin-like growth factor-1, or Ha-ras is
associated with activation of oocyte PDE and inhibition of adenylyl
cyclase. Although oocyte PDEs have not been characterized, specific
PDE3 inhibitors, but not PDE4 or PDE5 inhibitors, inhibit oocyte
maturation (54). It has been suggested (55, 56) that an insulin- and
cAMP-stimulated PDE3 in liver (phosphorylated and activated in
vitro by protein kinase A (56)) is important in the
antiglycogenolytic effects of insulin (41). In addition to PDE3, an
hepatocyte PDE4 is activated by insulin, suggesting that multiple PDEs
are involved in insulin signal transduction pathways (57).
In platelets, specific PDE3 inhibitors prevent aggregation, suggesting
that PDE3 is important in platelet function (58). Incubation of intact
platelets with insulin or effectors that increase cAMP, as well as
treatment in vitro with protein kinase A or a partially
purified PDE3IK, resulted in phosphorylation and activation of platelet
PDE3 (32, 59). As was found in adipocytes (30, 43), activation of the
platelet PDE3 by insulin correlates with serine phosphorylation (60)
although the effects of insulin activation of PDE3 on platelet function
are not known. Specific PDE3 inhibitors enhance myocardial
contractility and induce vascular and airway smooth muscle relaxation
(13-17). Both cardiac (33) and vascular smooth muscle (34) PDE3
isoforms are phosphorylated in vitro by protein kinase A;
little is known, however, about the regulation of these enzymes in
intact cells.
Many fundamental questions concerning PDE3 isoforms remain to be
addressed, including identification of the different PDE3 proteins in
cells and tissues, elucidation of mechanisms for regulation of their
gene expression, distribution, and activity, their structure-function relationships, and their role in the action of insulin and agents that
increase cAMP and cGMP. Perhaps more fundamental will be understanding
mechanisms for regulation of the overall PDE composition of individual
cells, functional integration of the different PDEs in establishing and
regulating cyclic nucleotide "fingerprints" of individual cells,
and the targeting of specific PDEs for therapeutic benefit.
-5
-phosphodiester bond of cyclic nucleotides, cyclic nucleotide
phosphodiesterases (PDEs)1 regulate
intracellular concentrations and effects of these second messengers.
PDEs include a large group of structurally related enzymes (reviewed in
Refs. 1-3). These enzymes belong to at least seven related gene
families (PDEs 1-7) (Fig. 1), which differ in their
primary structures, affinities for cAMP and cGMP, responses to specific
effectors, sensitivities to specific inhibitors, and mechanisms of
regulation (1-3). Most families are comprised of more than one gene;
14 different PDE genes have been identified. Within different families,
tissue-specific mRNAs are generated from the same gene by the use
of different transcription initiation sites or by alternative mRNA
splicing. Although some aspects of different PDE families will be
discussed, this review emphasizes the PDE3 family, including
structure-function information and regulation of the adipocyte PDE3,
which plays a key role in the antilipolytic action of insulin.
Fig. 1.
Structural organization of different PDE
families. Different genes within the same family are designated as
A, B, etc. Some family-specific inhibitors are in
brackets. The catalytic domain of PDE3s contains an
insertion of 44 amino acids () that does not align with other PDEs.
PDE6 can exist as a
2 heterotrimer of
catalytic and
inhibitory subunits;
and
subunits are products of different but closely related genes. This schematic does
not emphasize structural diversity found in N- and C-terminal regions
of variants in the different gene families, especially PDE1 and PDE4.
[P], location of potential phosphorylation sites. Regions
involved in calmodulin (CaM-BD) or cGMP binding
(cGMP-BD) are indicated. IBMX,
isobutylmethylxanthine.
[View Larger Version of this Image (36K GIF file)]
-inhibitory subunits of PDE6 by protein kinase C
alters their affinity for the PDE6
catalytic subunits. Effects of
phosphorylation on activities of PDEs 1, 5, and 6 in intact cells have
not been documented (1-3). A PDE4 isoform is phosphorylated and
activated in response to hormones that increase cAMP in intact cells
and by protein kinase A in vitro (1-3). Feedback regulation
of both PDE3 and PDE4 activities by cAMP-dependent
phosphorylation is likely to be central to intracellular mechanisms for
regulating the magnitude and duration of cAMP signals and responses and
desensitization to hormone signals (2).
Fig. 2.
Model of domain structure and membrane
association of PDE3. A, domain organization and deduced
amino acid (aa) identities of PDE3A and PDE3B isoforms.
Numbers are percentage of amino acid identity in the
indicated domain of the adjacent sequences (PC Gene, PALIGN).
B, membrane association of RPDE3B. The deduced sequence of
RPDE3B predicts five or six helical, potentially transmembrane, segments in the N-terminal hydrophobic region (PSORT, version 6.3; TM
pred (ISREC)). Based on the deduced sequence of RPDE3B, serine 302 within a protein kinase A consensus sequence (RRPS) is phosphorylated
in intact rat adipocytes incubated with insulin and/or isoproterenol.
The C-terminal catalytic domain is shown with the shaded
PDE3 insertion.
[View Larger Version of this Image (37K GIF file)]
Fig. 3.
Possible mechanisms for hormonal
regulation of the rat adipocyte PDE3B and its role in the antilipolytic
action of insulin. Stimulation of adipocytes with the
-adrenergic agonist/isoproterenol (ISO) or other
lipolytic agonists results in increased synthesis (via activation of
adenylyl cyclase (AC)) as well as increased degradation of
cAMP, the latter catalyzed by PDE3 and representing feedback regulation
of cAMP. The net increase in cAMP activates cAMP-dependent
protein kinase (A-kinase), which phosphorylates and
activates HSL, resulting in increased hydrolysis of triglyceride with
production of free fatty acids and glycerol. (Phosphorylation of HSL
might also be involved in its translocation to the surface of the
triglyceride storage droplet and increased access to substrate.) cAMP
activation of PDE3 correlates with phosphorylation of PDE3 (Ser-302)
and is presumably catalyzed by protein kinase A. The antilipolytic
action of insulin (INS) is initiated by activation of the
insulin receptor (IR) tyrosine protein kinase
(Tyr-Prk) activity. The autophosphorylated insulin receptor
phosphorylates IRS-1 on specific tyrosine residues, leading to a
recruitment of the regulatory (reg) subunit of PI3-K to
IRS-1 and activation of the catalytic subunit (cat) (44).
PI3-K could also be activated by direct interaction with insulin
receptor or by other signaling mechanisms independent of IRS-1 (61).
Activated PI3-K initiates a series of events leading to activation of
PDE3IK, which phosphorylates and activates PDE3B.
Phosphorylation/activation of PDE3 results in decreased cAMP/protein
kinase A activity, net dephosphorylation of HSL, and reduced lipolysis.
In the presence of insulin and
agonist, the condition during which
insulin normally exerts its antilipolytic effect, there is greater than
additive phosphorylation (Ser-302) and activation of PDE3B, consistent
with interaction between cAMP- and insulin-signaling pathways upstream
of PDE3B. PDE3IK has been partially characterized and is probably
itself activated by phosphorylation. Pretreatment of adipocytes with wortmannin, a PI3-K inhibitor, blocks activation of the PDE3IK, phosphorylation/activation of PDE3, and the antilipolytic effect of
insulin (42). Indirect or inferred pathways are indicated with
interrupted lines.
[View Larger Version of this Image (52K GIF file)]
-adrenergic receptors, adenylyl cyclase, protein kinase A,
Ca2+ channels, and PDEs (specific PDE isoforms were not
identified) were co-localized (52). Their results implicated PDEs as
critical regulators in limiting protein kinase A activation to
"local" Ca2+ channels and preventing diffusion of cAMP
to "distant" areas of the same cell. In human and rat pancreatic
islet preparations, PDE3, not PDE4, inhibitors stimulated insulin
secretion; this could reflect the presence of PDE3 and PDE4 in
different islet cells or different locations in the same cell (53).
We thank Dr. Martha Vaughan for critical review and Carol Kosh for expert secretarial assistance.