Cyclic 3′5′ adenosine monophosphate (cAMP) is a key second messenger that is responsible for regulating many pivotal signaling processes in all mammalian cell types (Tasken and Aandahl 2004). It influences cell growth, differentiation, shape, and movement as well as processes such a cardiac contraction, metabolism, water retention, learning, and memory. Most intriguingly, however, cAMP can selectively regulate a variety of very different processes in any one particular cell type. Indeed, uncovering the molecular mechanisms that allow cAMP to selectively regulate disparate processes within a single cell has provided a major challenge. Only very recently has the means whereby cAMP signaling is compartmentalized in cells begun to be understood (Tasken and Aandahl 2004; Baillie et al. 2005;Willoughby and Cooper 2007). One key requirement is for spatially discrete signaling complexes to be assembled in cells that have the machinery for detecting cAMP and translating this into an appropriate action. Critically, however, is a need for machinery able to dynamically regulate the access of cAMP to the detection system(s) and this is achieved by controlling the rate of cAMP degradation in the immediate vicinity of the signaling complex by targeted phosphodiesterases (Houslay 2010). Spatially constrained cAMP degradation has the functional consequence of adjusting the threshold for activation of cAMP sensors, in particular signaling complexes, in the face of a prevailing level of cAMP formation by adenylyl cyclase activity. Carrying out this process are forms of cAMP degrading phosphodiesterases (PDEs) that have specific amino acid sequence motifs, “zip codes”, allowing them to be targeted to particular signaling complexes (Houslay 2010). These PDEs convert cAMP to 5′-adenosine monophosphate (AMP) and so play a pivotal role in not only degrading the cAMP signal but, critically, in forming and shaping cAMP gradients in cells so as to regulate the activity of distinct signaling complexes. Indeed, recent technological advances have allowed cAMP gradients formed by PDEs to be dynamically visualized in living cells (Zaccolo 2009). These are FRET-based genetically encoded sensors that are based upon the cAMP effectors, either protein kinase A (PKA), which phosphorylates specific target proteins, or exchange protein activated by cAMP (EPAC), which activates the mini G-proteins, Rap1/2.
Our understanding of how targeted cAMP degradation underpins cAMP signaling has been revolutionized by studies on members of the multigene PDE4 subfamily.
What Is PDE4?
Long PDE4 isoforms possess dual regulatory domains that are unique to PDE4, namely, upstream conserved region 1 (UCR1) and upstream conserved region 2 (UCR2). These are located between the isoform-specific N-terminal region and the highly conserved catalytic unit (Fig. 1). UCR1 is tethered to UCR2 by the heterogeneous linker region LR1, with UCR2 tethered to the catalytic unit by the heterogeneous linker region, LR2. The PDE4 C-terminal regions from different subfamilies show no similarity but are identical for all active isoforms within any particular subfamily.
The presence or absence of UCR1/2 underpins classification; long forms have both, short forms lack UCR1, super-short forms lack UCR1 and have a truncated UCR2 while catalytically inactive dead-short forms lack both UCR1 and UCR2 and, through a unique 3′ splicing event, have a truncated catalytic unit and novel C-terminal region.
An important role of the UCR domains is to regulate quaternary structure where the UCR1/UCR2-conating long forms adopt a dimeric configuration, while the UCR1-lacking short and super-short forms are monomeric (Xie et al. 2014; Bolger et al. 2015).
Regulation by Phosphorylation
The rate of cAMP degradation by PDE4 can be dynamically regulated by phosphorylation. Pivotal to this are the UCR1 and UCR2 regulatory domains, which direct the functional outcome on the catalytic unit in response to phosphorylation (Houslay and Adams 2003; Mika and Conti 2016;Sheppard et al. 2014).
UCR1 contains a classical site for PKA phosphorylation, which confers activation on long isoforms. Thus, long isoforms play a prime role in the desensitization of cAMP signaling by facilitating cAMP degradation when cAMP levels rise sufficiently to activate PKA.
The catalytic unit of all PDE4 isoforms, save those from the PDE4A subfamily, contains a classical site for ERK phosphorylation. However, the UCR modules direct the functional outcome of such ERK phosphorylation, with long forms being inhibited by ERK phosphorylation, while short forms are activated and super-short isoforms show no activity change.
The inhibitory effect of ERK phosphorylation of PDE4 long isoforms can be overcome by PKA phosphorylation. This provides a potential feedback regulatory unit where ERK-mediated inhibition of a long PDE4 can cause local cAMP levels to rise, activating local PKA, which then phosphorylates the long PDE4 to disinhibit it, causing cAMP levels to fall. Thus, through this route, ligands activating ERK can lead to a programmed transient rise in cAMP through effects on PDE4 long forms.
UCR1 is also the site of phosphorylation by the stress-activated kinase, MK2 (MAPKAPK2) (MacKenzie et al. 2011). This does not alter PDE4 activity but attenuates the ability of PKA to activate long forms. An unidentified kinase activated by reactive oxygen species has also been shown to phosphorylate PDE4. Although this does not alter PDE4 activity, it reprograms the effect of inhibitory ERK phosphorylation of long isoforms to one of activation.
It has also been demonstrated that GSK3 can phosphorylate PDE4D isoforms, causing their ubiquitination and targeting for degradation.
Certain PDE4 forms are also modified by ubiquitination and SUMOylation. While neither alters PDE4 activity, SUMOylation of the PDE4D5 long isoform enhances its activation by PKA while reducing its inhibition by ERK phosphorylation.
Multisite phosphorylation confers dynamic regulation upon PDE4 activity and certain targeting events. In this, the UCR1 and UCR2 domains perform critical roles in defining the effect of phosphorylation on PDE4 catalytic activity. Unfortunately, there are no determined structures for full-length PDE4 isoforms due to the propensity of PDE4 enzymes to multimerize upon purification. However, biochemical studies have shown that UCR1 and UCR2 interact with each other, as does UCR2 with the catalytic unit. Indeed, a highly novel structural approach has demonstrated that UCR2 docks to the catalytic unit alongside the catalytic pocket (Cedervall et al. 2015). In fact, in certain conformational states a portion of UCR2 may even bind across the catalytic pocket, obscuring access to cAMP and engendering inhibition. This process may be regulated by phosphorylation, dimerization, and interaction with binding-partner proteins.
The core PDE4 catalytic unit is a compact structure of 17 α-helices folded into 3 subdomains. At the junction of the three subdomains is a deep cAMP substrate-binding pocket that contains two essential Me2+ binding sites, considered natively to be Zn2+ at the deeper site and Mg2+ at the more surface exposed site. The Zn2+ is tightly bound by four direct ligand interactions, with the fifth position being provided by water that bridges to the loosely held Mg2+ and plays the key role of nucleophilic attack in hydrolysis of the cyclophosphodiester bond. Structural changes arising from posttranslational modification, UCR1/2 domain interaction, and partner protein sequestration may be relayed into the catalytic center through helices 10 and 11.
Targeting of PDE4
Each PDE4 isoform has a signature N-terminal (N-Ter) domain whose major function is to confer intracellular targeting to specific signaling complexes and intracellular locations (Houslay 2010;Keravis and Lugnier 2010). The paradigm for this concept is the super-short PDE4A1 isoform that has a unique 25 amino acid N-Ter region. PDE4A1 is entirely membrane associated, being found in Golgi and vesicles that traffic from it. However, the engineered removal of its unique N-Ter region generates a soluble, cytosolic species that is clearly correctly folded as it remains fully catalytically active. This observation led to the notion that the PDE4 N-Ter region determined intracellular targeting, thus providing a functional reason for the diversity of PDE4 isoforms. Such targeting is invariably driven by protein–protein interactions although, as in the case of PDE4A1, it may also involve protein–lipid interactions.
All of the information required for membrane association and targeting is encompassed within the unique N-Ter region of PDE4A1 as when it is fused to various soluble proteins, the distribution of such chimeric species parallels that of full-length PDE4A1. NMR analysis of the PDE4A1 N-Ter region revealed two α-helical domains separated by a flexible hinge. The binding of Ca2+ to Asp21 in helix-2 elicits a conformational change that allows insertion of a core Trp19:Trp20 unit into the bilayer. However, PDE4A1 membrane targeting also requires helix-1, which increases the efficiency of helix-2 membrane association as well as conferring targeting to the trans-Golgi stack, probably by interacting with a Golgi-localized protein subsequent to membrane insertion of helix-1. Additionally, helix-1 allows PDE4A1 to undergo a dynamic intracellular redistribution to phosphatidic acid-rich regions upon Ca2+-binding to Asp6. Thus, PDE4A1 provides a paradigm for the targeting of PDE4 isoforms to specific complexes and for such interactions to be functionally, and dynamically, regulated.
Subsequently, specific PDE4 isoforms have been found to associate with signaling scaffold proteins (RACK1, myomegalin, β-arrestin, AKAPs, Ndel1, spectrin, Shank2, Lis1), receptors (p75 neurotropin receptor, ryanodine receptor, β1-adrenoceptor), enzymes (ERK, DNA-PKc), certain SH3 domain-containing proteins (e.g., Src family tyrosyl kinases), and transcription factors (AIP/XAP2). The mapping of interaction surfaces between many of these partnerships has been greatly facilitated by peptide array approaches. These have driven mutagenesis strategies to define interaction surfaces and in developing cell-permeable peptides for use in disrupting specific PDE4 partnerships in cells and so allowing functional assessment.
Dominant negative, siRNA, and gene-targeted knockout approaches have unequivocally shown that individual PDE4 isoforms can have specific functional roles and that there is no redundancy in the system. It is now appreciated that the diversity of PDE4 isoforms allows for targeting of particular species to specific, spatially defined signaling complexes in cells in order to allow for their exquisite control by gradients of cAMP sculpted by tethered PDE4 isoforms. A paradigm for this is the sequestration of the PDE4D5 isoform with the signaling scaffold β-arrestin. This complex is recruited to the β2-adrenergic receptor when cells are challenged with β-adrenergic receptor agonists, thus delivering an active cAMP degrading system to the site of cAMP synthesis. This complex has a dual desensitization role. Thus, β-arrestin uncouples the β2-adrenoceptor from the guanine nucleotide binding protein, Gs, and so negates activation of adenylyl cyclase, while PDE4D5 lowers local cAMP levels and switches off the activity of plasma membrane and β2-adrenergic receptor-tethered PKA. The importance of β-arrestin-delivered PDE4D5 in the control of these PKA subpopulations was gathered by a dominant-negative approach. This involved making a single point mutation deep within the cAMP-binding pocket of PDE4D5 so as to destroy catalytic activity while leaving the ability of PDE4D5 to bind β-arrestin unaltered. Overexpressing this inactive species in cells displaces active wild-type PDE4D5 from β-arrestin and allows plasma membrane PKA activity to rise unhindered in the face of challenge with β-adrenergic receptor agonists. However, no such action was seen if different catalytically inactive PDE4 isoforms were used or if a catalytically inactive PDE4D5 construct was engineered so as to disrupt its β-arrestin binding site.
Various of these PDE4 partnerships can be dynamically altered by external cues so as to reprogram cAMP gradients within cells. This can take the form of either changes in the expression of scaffolds, allowing new or competing partnerships to occur, or posttranslational modification that can alter specific partnerships. Thus the Mdm2-mediated ubiquitination of PDE4D5 enhances its interaction with β-arrestin while the PKA-mediated phosphorylation of PDE4D3 serves to enhance its interaction with the mAKAP scaffold while decreasing its interaction with the Ndel1 scaffold.
PDE4 and Disease
The first indication that PDE4 might provide an effective therapeutic target came from observations that rolipram, the first reported PDE4 selective inhibitor, exerted antidepressant properties. However, such therapeutic potential was negated by side effects, primarily severe emesis and nausea. Indeed, these issues have characterized, to a greater or lesser extent, all active-site-directed PDE4 inhibitors to date and provided a major challenge to the pharmaceutical industry in developing selective inhibitors with effective therapeutic windows (Houslay et al. 2005). PDE4 selective inhibitors also exert potent antiinflammatory actions and have been developed for treating inflammatory lung diseases such as asthma and COPD. Promisingly, in this regard, Roflumilast® has been approved for use in treating COPD in Europe and the USA.
The central problem with therapeutic exploitation of active-site-directed inhibitors is that they inhibit all PDE4 isoforms rather than just those that govern the target therapeutic process. As PDE4 catalytic units are identical in all isoforms from a particular subfamily and near identical between subfamilies, the chance of obtaining isoform-specific inhibitors through this route is nigh on impossible. Indeed generating subfamily selective inhibitors poses a severe challenge that has, to date, only shown scant possibilities for PDE4D and PDE4B. Further obfuscation arises from the fact that interacting proteins and posttranslational modification can alter the conformation of PDE4 isoforms. This, as shown in various instances, can translate into alterations in their susceptibility to active-site-directed inhibitors. Indeed, recent structural studies have given insight into how UCR2 interaction with the catalytic unit may drive some of these (Houslay and Adams 2010).
One possibility for the future is to determine whether the inhibition of a specific PDE4 isoform that was anchored could provide a route to treating defined diseases. If so, a novel therapeutic approach might be employed based upon developing either small molecules, peptides or peptidomimetics that disrupt such a key partnership. This would have the effect of displacing a specific PDE4 isoform from the site where it is needed to function in the cell, thereby raising only local cAMP levels, thus eliciting a highly targeted and specific action for therapeutic advantage while minimizing/negating side effects. Another would be to devise means of selectively inhibiting short forms rather than long forms as inhibition of the short PDE4B2 form underpins the anti-inflammatory response of PDE4 inhibition, while inhibition of long forms appears to be associated with various side-effects.
PDE4 diversity has been highly conserved in evolution. Interestingly, mutations in the PDE4 gene in Drosophila melanogaster have been described that lead to learning defects. This correlates with studies on mammals showing that cAMP is intimately linked with learning processes and a panoply of studies showing that PDE4 inhibitors can act as cognitive enhancers, which may offer future promise for the utility of PDE4 inhibitors in Alzheimer disease. More recently, the PDE4B gene has been linked to schizophrenia in certain families. Indeed, in an example of converging genetic and biochemical approaches, the scaffold protein DISC1, mutations in whose gene can provide strong association with schizophrenia, can interact with PDE4 isoforms, including PDE4B.
Mutations in the PDE4D gene have also been linked to stroke susceptibility. However, the linkage appears likely to depend on both the type of stroke and ethnic background. The basis of the linkage is obscure to date but mutations appear to locate in noncoding regions upstream of the PDE4D7 isoform, which may indicate effects on the specific promoter for this isoform.
The specific inhibitory PDE4D mutations have been shown to lead to the molecular pathology of acrodysostosis without hormone resistance. However, interestingly, it would appear that the pathological phenotype may well be dependent on an overcompensatory induction of other PDE4 isoforms that can be expected to be targeted to different signaling complexes and exert distinct effects on compartmentalized cAMP signaling (Kaname et al. 2014).
Altering PDE4-related cAMP signaling in medium spiny neurons of the ventral striatum has been shown to modulate stress-induced behavioral states. This occurs through the Cdk5-mediated phosphorylation of PDE4 and interdicting this regulatory pathway could provide a new therapeutic approach for clinical conditions associated with stress, such as depression (Plattner et al. 2015).
Brief periods of sleep loss have been shown to have long-lasting consequences related to impaired memory consolidation, which appears to be due to upregulation of cAMP-degrading PDE4A4/PDE4A5 activity. This occurs through a cAMP-PDE4-PKA-LIMK-cofilin activation-signaling pathway (Havekes et al. 2016).
PDE4 inhibitors have also been suggested to have therapeutic possibilities for treating certain cancers. They, seemingly, can promote B-cell apoptosis in chronic lymphocytic leukemia and can inhibit the growth and migration of colon cancer cells, for example (Serrels et al. 2010).
PDE4 isoforms provide a major route for underpinning compartmentalized cAMP signaling in all mammalian cells. The plethora of isoforms provides a library of species that can be targeted to specific intracellular locales/signaling complexes. This enables them to sculpt local cAMP gradients, thereby gating the activation of tethered cAMP effectors to cAMP. The selective expression of PDE4 isoforms with different regulatory regions (UCR1/2) allows them to perform distinct roles in both the cellular processes for desensitizing cAMP signaling and for crosstalk with other critical signaling systems so as to integrate cellular responses in cell-type specific fashions.
Members of the diverse PDE4 family are invariably located at nodes that link multiple signaling pathways and, as such, are set to play key roles in regulating pivotal cellular responses. Our understanding of the range of PDE4 partner proteins and multiple sites for phosphorylation, ubiquitination, and SUMOylation is only just beginning to be appreciated. Furthermore, there is also the exciting possibility that the function of the PDE4 enzymes is not just to carry out targeted cAMP degradation but also to provide a signaling role for themselves (Murdoch et al. 2011).
Current studies have served to set the scene for a more detailed appreciation of the importance and pervasiveness of PDE4 isoforms, their key roles in biology, health and disease, and future therapeutic exploitation.
- Cheung YF, Kan Z, Garrett-Engele P, Gall I, Murdoch H, Baillie GS, Camargo LM, Johnson JM, Houslay MD, Castle JC. PDE4B5, a novel, super-short, brain-specific cAMP phosphodiesterase-4 variant whose isoform-specifying N-terminal region is identical to that of cAMP phosphodiesterase-4D6 (PDE4D6). J Pharmacol Exp Ther. 2007;322:600–9.PubMedCrossRefGoogle Scholar
- Havekes R, Park AJ, Tolentino RE, Bruinenberg VM, Tudor JC, Lee Y, Hansen RT, Guercio LA, Linton E, Neves-Zaph SR, Meerlo P, Baillie GS, Houslay MD, Abel T. Compartmentalized PDE4A5 Signaling Impairs Hippocampal Synaptic Plasticity and Long-Term Memory. J Neurosci. 2016;36(34):8936–46. doi:10.1523/JNEUROSCI.0248-16.2016.Google Scholar
- Kaname T, Ki CS, Niikawa N, Baillie GS, Day JP, Yamamura K, Ohta T, Nishimura G, Mastuura N, Kim OH, Sohn YB, Kim HW, Cho SY, Ko AR, Lee JY, Kim HW, Ryu SH, Rhee H, Yang KS, Joo K, Lee J, Kim CH, Cho KH, Kim D, Yanagi K, Naritomi K, Yoshiura K, Kondoh T, Nii E, Tonoki H, Houslay MD, Jin DK. Heterozygous mutations in cyclic AMP phosphodiesterase-4D (PDE4D) and protein kinase A (PKA) provide new insights into the molecular pathology of acrodysostosis. Cell Signal. 2014;26:2446–59.PubMedCrossRefGoogle Scholar
- Keravis T, Lugnier C. Cyclic nucleotide phosphodiesterase (PDE) isozymes as targets of the intracellular signalling network: benefits of PDE inhibitors in various diseases and perspectives for future therapeutic developments. Br J Pharmacol. 2012;165:1288–305.PubMedPubMedCentralCrossRefGoogle Scholar
- MacKenzie KF, Wallace DA, Hill EV, Anthony DF, Henderson DJ, Houslay DM, Arthur JS, Baillie GS, Houslay MD. Phosphorylation of cAMP-specific PDE4A5 (phosphodiesterase-4A5) by MK2 (MAPKAPK2) attenuates its activation through protein kinase A phosphorylation. Biochem J. 2011;435(3):755–69. doi:10.1042/BJ20101184.Google Scholar
- Plattner F, Hayashi K, Hernández A, Benavides DR, Tassin TC, Tan C, Day J, Fina MW, Yuen EY, Yan Z, Goldberg MS, Nairn AC, Greengard P, Nestler EJ, Taussig R, Nishi A, Houslay MD, Bibb JA. The role of ventral striatal cAMP signaling in stress-induced behaviors. Nat Neurosci. 2015;18:1094–100.PubMedPubMedCentralCrossRefGoogle Scholar
- Serrels B, Sandilands E, Serrels A, Baillie G, Houslay MD, Brunton VG, Canel M, Machesky LM, Anderson KI, Frame MC. A complex between FAK, RACK1, and PDE4D5 controls spreading initiation and cancer cell polarity. Curr Biol. 2010;20(12):1086–92. doi:10.1016/j.cub.2010.04.042.Google Scholar
- Sheppard CL, Lee LC, Hill EV, Henderson DJ, Anthony DF, Houslay DM, Yalla KC, Cairns LS, Dunlop AJ, Baillie GS, Huston E, Houslay MD. Mitotic activation of the DISC1-inducible cyclic AMP phosphodiesterase-4D9 (PDE4D9), through multi-site phosphorylation, influences cell cycle progression. Cell Signal. 2014;26:1958–74.PubMedCrossRefGoogle Scholar