Role of phosphodiesterases in the shaping of sub-plasma-membrane cAMP oscillations and pulsatile insulin secretion

Cyclic AMP (cAMP) is a ubiquitous intracellular messenger, which regulates numerous cell functions, including metabolism, gene expression and secretion. Changes of cAMP concentration are typically spatially confined and precisely regulated in time. The tight spatio–temporal control is an important prerequisite for the versatility and specificity of cAMP signalling pathways, and is, at least in part, obtained by subcellular compartmentalization of adenylyl cyclases and phosphodiesterases (PDEs), the enzymes generating and degrading the nucleotide, respectively (Conti and Beavo, 2007; Willoughby and Cooper, 2007; Houslay, 2010). The subcellular localisation of adenylyl cyclases and PDEs is determined by protein–protein and protein–lipid interactions, often through anchoring proteins that also bind regulating factors and downstream effectors, such as protein kinases, phosphatases and their substrates, thereby generating spatially compartmentalized signalling complexes (Taskén and Aandahl, 2004; Wong and Scott, 2004).

The PDEs constitute a large group of enzymes that are divided into 11 subfamilies, which differ with regard to substrate specificity (some of them hydrolyzing cyclic GMP), kinetics and regulatory properties (Bender and Beavo, 2006; Lugnier, 2006; Conti and Beavo, 2007). The PDE1 family of enzymes is activated by Ca²⁺-calmodulin (Goraya and Cooper, 2005), and cAMP hydrolysis by PDE2 is stimulated by cGMP. PDE3 is a dual-specificity enzyme with kinetic properties resulting in cGMP-inhibited cAMP degradation (Thompson et al., 2007; Murata et al., 2009). PDE4, PDE7 and PDE8 are all cAMP-specific enzymes. The more than 20 different PDE4 family members are characterised by multiple different molecular targeting interactions for subcellular localisation as well as by regulation by phosphorylation mediated through protein kinase A (PKA), extracellular signal-regulated kinase (ERK) and MAP-kinase-activated protein kinase 2 (MK2) (Houslay and Adams, 2003; Houslay, 2010). PDE5, PDE6 and PDE9 are cGMP-selective, and the relatively recently described and yet poorly characterised PDE10 and 11 family enzymes show dual substrate specificity (Bender and Beavo, 2006).

PDE activity has been extensively characterised in many cell types by using biochemical methods based on hydrolysis of radioactive nucleotides. All isoforms, except for PDE8 and PDE9, can be inhibited by 3-isobutyl-methylxanthine (IBMX) (Lugnier, 2006). The recent development of tools for online monitoring of cAMP concentration (Zaccolo et al., 2000; Rich et al., 2001; Zaccolo and Pozzan, 2002; DiPilato et al., 2004; Nikolaev et al., 2004; Dyachok et al., 2006a; Herget et al., 2008) or PDE activity (Takakusa et al., 2002) in living cells has enabled more detailed studies of the function of PDEs in compartmentalization of cAMP signalling. For example, cAMP measurements with cyclic nucleotide gated channels or fluorescence sensors have shown that the cAMP concentration displays varying kinetics in different subcellular compartments owing to compartmentalization of PDEs (Rich et al., 2001; Terrin et al., 2006). Real-time cAMP measurements have also revealed complex temporal patterns, including oscillations. cAMP oscillations were first described as an important property of developing Dictyostelium slime moulds (Gerisch and Wick, 1975). These oscillations involve the excretion of cAMP with binding to extracellular cell surface receptors (Maeda et al., 2004). cAMP oscillations have subsequently been demonstrated to occur spontaneously in developing neural circuits (Gorbunova and Spitzer, 2002; Dunn et al., 2006) and after treatment of insulin-secreting MIN6 cells with a combination of high concentrations of glucose and a K⁺ channel inhibitor (Landa et al., 2005). We recently developed a ratiometric evanescent wave microscopy approach to determine the concentration of cAMP in the sub-plasma-membrane space ([cAMP]_pm) and found that [cAMP]_pm undergoes pronounced oscillations in isolated hormone- (Dyachok et al., 2006a) and glucose-stimulated (Dyachok et al., 2008) insulin-secreting cells as well as in α- and β-cells within intact pancreatic islets (Tian et al., 2011). However, the mechanisms underlying these oscillations are poorly understood.

In pancreatic β-cells, cAMP strongly amplifies the secretion of insulin. The nucleotide acts through both PKA and the cAMP-dependent guanine nucleotide exchange factor EPAC at multiple levels to potentiate Ca²⁺-dependent exocytosis (Ämmälä et al., 1993; Renström et al., 1997; Dyachok and Gylfe, 2004; Seino and Shibasaki, 2005). The coordination of Ca²⁺ and cAMP oscillations are important to generate pulsatile insulin secretion (Dyachok et al., 2008). Pancreatic islets and insulin-secreting β-cells express several PDE isozymes. Early studies demonstrated that pancreatic islets show Ca²⁺-calmodulin-sensitive PDE activity (Sugden and Ashcroft, 1981; Lipson and Oldham, 1983; Capito et al., 1986), and expression of PDE1 isoforms has subsequently been demonstrated by RT-PCR analysis in islets and β-cell lines (Han et al., 1999; Dov et al., 2008; Waddleton et al., 2008). PDE3 exists in two isoforms, PDE3A and PDE3B. PDE3B is expressed in β-cells and has been demonstrated to be important for insulin secretion (Parker et al., 1995; Zhao et al., 1997; Härndahl et al., 2002; Pyne and Furman, 2003; Choi et al., 2006; Waddleton et al., 2008), probably by affecting the most distal steps of granule fusion (Walz et al., 2007). Furthermore, PDE4 has been found to have a function in islets and insulin-secreting cells (Parker et al., 1995; Shafiee-Nick et al., 1995; Waddleton et al., 2008). Recent studies have also identified members of the PDE5, PDE7, PDE8, PDE9, PDE10 and PDE11 families in rodent and human islets and in insulin-secreting cell lines (Waddleton et al., 2008; Heimann et al., 2010). Silencing of PDE8B expression with RNA interference was found to potentiate insulin secretion in rat islets and insulinoma cells (Dov et al., 2008; Waddleton et al., 2008). An increasing number of family-selective pharmacological PDE inhibitors have been reported, and PDEs are attractive targets for enhanced insulin secretion in type 2 diabetes. The aim of the present study was to determine how different PDE families contribute to the regulation of cAMP concentration in the sub-plasma-membrane space and, in particular, their functions in the generation of the pronounced [cAMP]_pm oscillations in β-cells that underlie the glucose-stimulated pulsatile release of insulin.

In β-cells exposed to a basal medium containing 3 mM glucose, the [cAMP]_pm was low and stable. Inhibition of PDEs with IBMX invariably resulted in an increase in [cAMP]_pm. At a concentration of 50 µM, IBMX caused a stable increase in [cAMP]_pm in 11 out of 20 MIN6 cells (55%, Fig. 1A). In the remaining cells, the [cAMP]_pm increase was either transient with return to baseline levels after 5–10 minutes, or showed oscillations typically occurring from an elevated level (Fig. 1B). At 300 µM, IBMX induced a stable increase in [cAMP]_pm in 29 out of 36 cells (81%) and oscillations in the remaining seven cells (not shown). Although there was no difference in the apparent [cAMP]_pm level reached after stimulation with IBMX (50 µM, 1.46±0.07 ratio units; 300 µM, 1.46±0.05 ratio units), the maximal rate of increase in [cAMP]_pm was more than doubled in the presence of 300 µM IBMX (Fig. 1C). Similar results were obtained in primary mouse β-cells. Thus, 50 µM IBMX induced a stable increase in [cAMP]_pm in 13 out of 18 cells (72%, Fig. 1D) and oscillations in the remaining cells (Fig. 1E). The IBMX-induced [cAMP]_pm increase was reversed by 5 µM adrenaline, which suppresses the formation of cAMP in β-cells by activating G_i-coupled α2-adrenoceptors (Schuit and Pipeleers, 1986). This response can be used to distinguish β-cells from α-cells, in which adrenaline increases [cAMP]_pm by activating G_s-coupled β-adrenoceptors (Schuit and Pipeleers, 1986; Tian et al., 2011) (Fig. 1D,E). The oscillatory [cAMP]_pm responses probably reflect incomplete inhibition of PDEs in combination with natural variations in the rate of cAMP production. Analysis of the IBMX-induced increase in [cAMP]_pm revealed that it sometimes occurred in two steps, with a rapid rise that levelled off before increasing again to a steady-state level (Fig. 1D,F). In other cases, the opposite pattern was observed, that is, a slow initial response followed by a sudden acceleration (data not shown). These biphasic responses were seen in 55% (n = 20) of MIN6 cells and 50% (n = 18) of primary β-cells at 50 µM IBMX, and in 30.5% (n = 36) of MIN6 cells at 300 µM of the PDE inhibitor. Periodic changes of [cAMP]_pm were never observed when cAMP production was suppressed. Thus, inhibition of membrane adenylyl cyclase activity with 100 µM 2′,5′-dideoxyadenosine (DDA) resulted in a mono-exponential reduction in [cAMP]_pm without oscillations to a level corresponding to 36±4% (n = 6) or 56±4% (n = 17) of that obtained with 50 µM (data not shown) or 300 µM IBMX (Fig. 1G). A similar reduction in [cAMP]_pm (to 61±5%; n = 9) was obtained even in the presence of 1 mM of the PDE inhibitor (Fig. 1H). Together, these data indicate that constitutive PDE activity masks periodic basal levels of cAMP production beneath the plasma membrane and that the cAMP degradation is accounted for both by IBMX-sensitive and -insensitive mechanisms.

To determine the PDE isoforms that maintain low [cAMP]_pm under basal (i.e. non-stimulated) conditions, PDE family-specific pharmacological inhibitors were applied to MIN6 cells or islets in the presence of 3 mM glucose. In MIN6 cells, inhibition of isoforms from the PDE4 family with 8 µM rolipram caused only a very modest [cAMP]_pm elevation in 11% of the cells (n = 44; Fig. 2A,K,L). A larger fraction of the cells (33%, n = 15) responded to 100 µM of the PDE1 inhibitor 8-methoxymethyl-IBMX (MM-IBMX), showing a CFP:YFP ratio increase of 0.13±0.03 (n = 5; Fig. 2B,K,L). The PDE3 inhibitor cilostamide (0.7–1 µM) caused an increase in [cAMP]_pm in 52% of the cells (n = 29; ratio increase 0.16±0.04; Fig. 2C,K,L). Increasing the concentration to 8 µM did not increase the number of responding cells, but at this concentration, cilostamide induced oscillations of [cAMP]_pm in five out of 12 cells (Fig. 2D). Similar results were obtained with primary β-cells within mouse islets, but compared with MIN6 cells, the primary cells were more responsive to PDE4 inhibition (Fig. 2E–G; Fig. 2K,L). The inhibitors of PDE4, PDE1 and PDE3 thus induced an increase in [cAMP]_pm in 39% (five out of 13 cells; 0.16±0.014 ratio units; Fig. 2E,K,L), 27% (three out of 11 cells; 0.19±0.009 ratio units; Fig. 2F,K,L) and 50% (five out of 10 cells; 0.23±0.03 ratio units; Fig. 2G,K,L), respectively, of β-cells identified with 5 µM adrenaline.

The lack of response to inhibitors of a single PDE family in many cells might be owing to other isoforms showing sufficiently high activity to maintain low [cAMP]_pm. Accordingly, when MM-IBMX and cilostamide were combined, there was a response in 95% of the MIN6 cells with an average CFP:YFP increase of 0.34±0.04 ratio units (n = 19; Fig. 2H,K,L). Addition of IBMX increased the [cAMP]_pm by an additional 0.10±0.03 units. In a similar manner, the combination of cilostamide and rolipram evoked a response in 79% of the cells, averaging 0.31±0.03 ratio units (n = 21; Fig. 2I–L). These findings indicate that PDE3 accounts for most of the basal levels of IBMX-sensitive PDE activity in the sub-membrane space of β-cells. Although PDE1 is the second most important isoform in MIN6 cells, PDE4 and PDE1 both contribute substantially in primary mouse β-cells.

Next, we investigated which PDE isoforms are involved to shape the [cAMP]_pm signals in glucose-stimulated cells. As previously demonstrated in our laboratory, for individual β-cells (Dyachok et al., 2008) and β-cells within intact islets (Tian et al., 2011), elevation of the glucose concentration often results in pronounced oscillations of [cAMP]_pm. The amplitudes of the oscillations in MIN6 cells averaged 0.27±0.02 ratio units and the period was 4.7±0.4 minutes (Fig. 3A; n = 21). In contrast to its minimal effect in non-stimulated MIN6 cells, the PDE4 inhibitor rolipram induced a transient increase in [cAMP]_pm in cells exposed to 20 mM glucose (Fig. 3B; n = 7). After the first peak, [cAMP]_pm oscillations typically continued unaltered in the presence of the PDE4 inhibitor (Fig. 3B,G). In cells showing an initial glucose-induced [cAMP]_pm transient followed by a stable increase at suprabasal levels of [cAMP]_pm, rolipram sometimes evoked a sustained oscillatory pattern (Fig. 3C, n = 6). Overall, rolipram significantly increased the time-average level of [cAMP]_pm (Fig. 3H). Primary mouse islet β-cells responded in a similar way to the MIN6 cells, and in the presence of 20 mM glucose, rolipram caused a modest amplification of the sugar-induced [cAMP]_pm oscillations (Fig. 3D; n = 4) or evoked pronounced oscillations from a slightly elevated baseline level (Fig. 3E; n = 2). Although PDE4 activity is known to be regulated by PKA (Houslay and Adams, 2003), the rolipram-induced elevation of [cAMP]_pm in glucose-stimulated MIN6 cells was not affected by 1–5 µM of the PKA inhibitor H89 (0.14±0.03, n = 15 with 5 µM H89 vs 0.14±0.02, n = 16 in control; Fig. 3F). Thus, PDE4 seems to be activated by glucose stimulation in a PKA-independent manner in insulin-secreting cells, but is not required for the generation of [cAMP]_pm oscillations.

The majority of MIN6 β-cells stimulated with 20 mM glucose responded to 0.7 µM of the PDE3 inhibitor cilostamide with an increase in [cAMP]_pm (83%, n = 24), and in 50% of the responding cells there were oscillations in [cAMP]_pm (Fig. 4A). The frequency of these oscillations did not differ from the control, but the time-integrated [cAMP]_pm response was significantly amplified (Fig. 4G,H). Similar responses were seen with 8 µM cilostamide, with some cells showing continued oscillations from an elevated level (n = 12; Fig. 4B), and others showing a stable increase in [cAMP]_pm (n = 11; not shown). Primary mouse islet β-cells responded essentially as MIN6 cells. Thus, five out of eight cells responded to 1 µM cilostamide with an increase in [cAMP]_pm and continued oscillations (Fig. 4C).

In the presence of 20 mM glucose, the responses to the PDE1 inhibitor MM-IBMX were similar to those of cilostamide in both MIN6 and primary mouse islet β-cells. Accordingly, most of the cells responded with an increase in [cAMP]_pm (18 of 24 MIN6 cells, and six out of eight primary β-cells). In some MIN6 cells, the glucose-induced [cAMP]_pm oscillations were perturbed by the drug (n = 5; Fig. 4D), but in other cases the oscillations continued (n = 4) or were even restored by MM-IBMX with maintained frequency and an increased average [cAMP]_pm level (n = 8; Fig. 4E,G,H). An example of a primary mouse islet β-cell where MM-IBMX increases the average [cAMP]_pm level without inhibiting the oscillatory pattern is shown in Fig. 4F. Exposure to IBMX induced further [cAMP]_pm elevation, which was counteracted by adrenaline. These results demonstrate that PDE1 and PDE3 contribute to shaping the glucose-induced [cAMP]_pm signals, but neither of them are essential for the generation of cAMP oscillations.

We next assessed the involvement of IBMX-insensitive PDEs in shaping [cAMP]_pm signals in β-cells. Measurements of PDE activity in MIN6 β-cell lysates showed that 7.2±0.3% (n = 5) of the total PDE activity remained after inhibition of PDE1, PDE3 and PDE4 with MM-IBMX, cilostamide and rolipram. IBMX inhibited an additional 54±6% showing that the IBMX-insensitive activity only constitutes a few percent of the total PDE activity. Nevertheless, IBMX-insensitive PDEs might be functionally important in specific subcellular compartments. Among IBMX-insensitive isoforms, PDE8B is expressed in islets and insulin-secreting cell lines (Dov et al., 2008; Waddleton et al., 2008). The limited availability of useful pharmacological inhibitors of PDE8 makes it difficult to study this family of enzymes. Dipyridamole can be used to inhibit PDE8, but the drug is fluorescent and interfered spectrally with the present [cAMP]_pm biosensor measurements (data not shown). In the PDE assay, dipyridamole inhibited 63±3% (n = 5) of the small fraction remaining after inhibition of PDE1, PDE3 and PDE4. The non-specific PDE inhibitor papaverine (100 µM) induced increases in [cAMP]_pm in the presence of IBMX and reduced the [cAMP]_pm-lowering effect of adrenaline, consistent with the presence of IBMX-insensitive PDE activity in primary mouse β-cells (Fig. 5A). To pinpoint the involvement of enzymes of the PDE8 family, MIN6 cells were treated with lentiviral vectors expressing shRNA directed against PDE8B. After 48 hours, the PDE8B mRNA level was approximately half of that in control cells expressing non-target shRNA, whereas the levels of PDE1C, PDE3B and PDE4A were unaffected (Fig. 5B). Knockdown of PDE8B did not result in any significant change of the small dipyridamole-sensitive or IBMX-insensitive PDE activity in the cell lysates (data not shown). However, in the single-cell measurements, the basal level of [cAMP]_pm was elevated in knockdown cells as indicated by a distinct and reversible [cAMP]_pm-lowering effect of adenylyl cyclase inhibition with DDA (100 µM), which was not observed in control cells (Fig. 5C).

The glucose response in MIN6 cells treated with non-target shRNA lentivirus was unaffected, and 25 out of 29 cells (86%) reacted to a step increase of glucose from 3 to 20 mM with pronounced [cAMP]_pm oscillations (Fig. 5D). Elevation of [cAMP]_pm induced by 50 µM IBMX was almost completely suppressed by DDA (Fig. 5D,F). After knockdown of PDE8B, the increase in [cAMP]_pm in response to glucose was less pronounced (Fig. 5E,F) and only 45% of the cells (n = 31) showed oscillations, which in many cases occurred from an elevated level. In 32% of the cells, the glucose-induced increase in [cAMP]_pm was stable. IBMX induced a higher response in knockdown cells compared with that of the control (Fig. 5F) and 100 µM DDA failed to reduce the level of [cAMP]_pm in the presence of IBMX (Fig. 5E,F). These data indicate that PDE8B accounts for most of the IBMX-insensitive PDE activity in the β-cell sub-plasma-membrane space, and that this isoform has an important function both for determining the basal level of [cAMP]_pm and for shaping glucose-induced [cAMP]_pm oscillations.

Because [cAMP]_pm oscillations are important for the magnitude and kinetics of insulin secretion (Dyachok et al., 2008), we investigated the effect of the various PDEs on insulin-release kinetics. The time-course of insulin secretion from single MIN6 cells was monitored using the fluorescent translocation sensor GFP₄–GRP1, which reports formation of PtdIns(3,4,5)P₃ in the plasma membrane following insulin receptor activation (Dyachok et al., 2008). Most cells that express the sensor responded to a rise of glucose from 3 to 20 mM with pronounced PtdIns(3,4,5)P₃ oscillations reflecting pulsatile secretion of insulin (46 of 54 cells; Fig. 6A,F,G). Inhibition of PDE3 with cilostamide did not affect the frequency of oscillations but caused a modest, though significant, increase in the peak amplitude of the oscillations and a marked increase in the time-average levels of PtdIns(3,4,5)P₃ (Fig. 6A; Fig. 6F–H). Removal of cilostamide was sometimes associated with reduction of PtdIns(3,4,5)P₃ and loss of the oscillations, with restoration of the response only occurring after reintroduction of the drug (Fig. 6A). In eight of the 54 cells, glucose triggered an initial increase in the level of PtdIns(3,4,5)P₃ followed by a modest stable elevation. Cilostamide induced oscillations of PtdIns(3,4,5)P₃ from an elevated level (data not shown). Rolipram affected neither the amplitude nor the frequency of the PtdIns(3,4,5)P₃ oscillations, but in a similar manner to cilostamide, removal of the drug sometimes resulted in loss of the oscillatory response (Fig. 6B; Fig. 6F–H; n = 47).

Inhibition of PDE1 with 100 µM MM-IBMX resulted in a marked increase in the level of PtdIns(3,4,5)P₃ and loss of the oscillations (n = 39; Fig. 6C,H). A similar perturbation of the oscillatory response was obtained with 30 µM of the drug, but this was not associated with elevation of the time-average level of PtdIns(3,4,5)P₃ (Fig. 6H; n = 61). Similarly, after PDE8B knockdown, the glucose-induced PtdIns(3,4,5)P₃ response was typically stable (Fig. 6D,E; 25 of 27 cells) and the time-average GFP₄–GRP1 fluorescence was nearly doubled compared with the control (Fig. 6H), consistent with a marked amplification of insulin secretion.

The specificity and versatility in cAMP signalling pathways depend on spatial compartmentalization and precise temporal control of the cAMP concentration. Pancreatic β-cells constitute an interesting model to study the spatio–temporal dynamics of cAMP because these cells display pronounced cAMP oscillations in the sub-membrane space after stimulation with physiological regulators of insulin secretion (Dyachok et al., 2006a; Dyachok et al., 2008). In the present study, we investigated the involvement of PDEs in shaping sub-membrane cAMP signals and found that no single PDE isoform is essential for the generation of cAMP oscillations, but that [cAMP]_pm depends on periodic cAMP production in combination with constitutive cAMP degradation through both IBMX-sensitive and -insensitive mechanisms. PDE8B was found to account for most of the IBMX-insensitive PDE activity in the sub-membrane space, and together with PDE1 and PDE3, has a key function in the regulation of insulin-secretion kinetics.

Inhibition of most PDEs by IBMX resulted in dose-dependent elevation of [cAMP]_pm at 3 mM glucose, consistent with cAMP production being balanced by PDE-mediated degradation. Although IBMX is also an adenosine receptor inhibitor, and β-cells express adenosine A₁ receptors, the activation of which results in a reduction in cAMP (Bertrand et al., 1989), it is unlikely that A₁-receptor antagonism mediated by IBMX would contribute to the [cAMP]_pm elevation in unstimulated cells. The observation that [cAMP]_pm often fluctuates in the presence of low IBMX concentrations indicates that there are variations in the basal rate of cAMP production that become unmasked upon partial suppression of cAMP degradation. As expected, further suppression of cAMP degradation resulted in the disappearance of the oscillations. Variations in the rate of cAMP production should also be reflected in the rate of [cAMP]_pm elevation. Indeed, the IBMX-induced increase of [cAMP]_pm was sometimes delayed or occurred in two steps. The temporary ‘plateau’ at an intermediate [cAMP]_pm level might thus reflect a period of relatively low cAMP production.

Periodic cAMP production could result from variations in metabolism. It is well established that metabolism in β-cells and other types of cells oscillates (Hess and Boiteux, 1971; Longo et al., 1991; Jung et al., 2000). We have recently shown that glucose-induced cAMP production is mediated by an increase in the levels of metabolically derived ATP (Dyachok et al., 2008). Because ATP seems to oscillate in both glucose-stimulated β-cells (Ainscow and Rutter, 2002) and under basal conditions (Dryselius et al., 1994), it is conceivable that the oscillations in [cAMP]_pm reflect those of ATP. Apart from being a precursor for cAMP, it remains to be established if ATP also has other regulating influences on adenylyl cyclases. The possibility that oscillations in [cAMP]_pm instead are due to inherent variations in PDE activity appears unlikely. The presence of [cAMP]_pm oscillations when most PDEs were inhibited by IBMX does not exclude the fact that IBMX-insensitive mechanisms show periodic activity. However, [cAMP]_pm oscillations were not observed after inhibition of cAMP production, which would have been expected if PDE activity showed intrinsic oscillations.

Pharmacological inhibitors were used to identify PDE families involved in cAMP degradation under basal and glucose-stimulated conditions. MIN6 cells and β-cells within intact mouse islets overall responded in a similar manner. The effects of the PDE1 inhibitor MM-IBMX were modest in the presence of 3 mM glucose, which does not exclude that this enzyme is active under basal conditions. An increase in cAMP might escape detection if it occurs outside the sub-membrane compartment where the biosensor is located. The PDE activity that remains when a single isoform is inhibited might still be sufficient to maintain low levels of [cAMP]_pm. Consistent with the latter idea, it was found that 95% of the cells responded to a combination of MM-IBMX and the PDE3 inhibitor cilostamide, whereas only 33 and 52%, respectively, responded to either of the drugs alone. A similar synergistic effect was observed with combined inhibition of PDE3 and PDE4. The MM-IBMX response was more pronounced in glucose-stimulated cells, which is not surprising, because glucose stimulation of β-cells is associated with elevation of the cytoplasmic Ca²⁺ concentration (Grapengiesser et al., 1988) and PDE1 is activated by Ca²⁺-calmodulin (Goraya and Cooper, 2005). As the glucose-induced [cAMP]_pm oscillations continued in many cells exposed to MM-IBMX, it seems unlikely that Ca²⁺ regulation of PDE1 underlies the oscillations as has been suggested for MIN6 cells stimulated with a combination of glucose and the K⁺-channel inhibitor tetraethylammonium (Landa et al., 2005). Their independence of the Ca²⁺-calmodulin-sensitive PDE1 is in line with our previous observation that the oscillations often persist when Ca²⁺ entry is prevented (Dyachok et al., 2008). Nevertheless, consistent with previous studies (Han et al., 1999; Waddleton et al., 2008) inhibition of PDE1 markedly amplified insulin release, and this effect was associated with a disturbed pulsatile secretion pattern.

The rise of [cAMP]_pm after inhibition of PDE4 in primary β-cells is consistent with the previously reported role of this isoform in islets (Parker et al., 1995; Shafiee-Nick et al., 1995; Waddleton et al., 2008). In the MIN6 cells, PDE4 seemed active only after stimulation with glucose. The mechanism underlying the activation of PDE4 by glucose is unclear. The enzyme shows a relatively low affinity for cAMP (Bender and Beavo, 2006), and activity is therefore stimulated by the elevated levels of the nucleotide. Although some PDE4 isoforms are regulated by PKA (Houslay and Adams, 2003), the glucose-induced PDE4 activity did not depend on this kinase, as rolipram also readily increased [cAMP]_pm in the presence of H89. Inhibition of PDE3 gave the most pronounced increases of [cAMP]_pm in both resting and glucose-stimulated cells, which is consistent with previous studies identifying PDE3B as a main regulator of cAMP in cellular compartments that are relevant for insulin secretion. For example, β-cells or transgenic mice overexpressing PDE3B show reduced insulin secretion (Härndahl et al., 2002; Härndahl et al., 2004), and genetic or pharmacological reduction of PDE3B activity amplifies secretion (Choi et al., 2006; Waddleton et al., 2008). The strong effect of a PDE3 inhibitor on the concentration of cAMP in the sub-membrane space is also in line with the observation that PDE3 is often associated with membranes (Shakur et al., 2001), including the plasma membrane and secretory-granule membranes (Walz et al., 2007). Although [cAMP]_pm levels and insulin release were greatly increased by cilostamide, the drug neither prevented [cAMP]_pm oscillations nor interfered with pulsatile insulin secretion.

Previous biochemical studies have indicated that 90% of total PDE activity in insulin-secreting cells can be inhibited by IBMX (Pyne and Furman, 2003), and we have now found that an even higher percentage can be inhibited in MIN6 β-cells. However, this does not mean that the remaining fraction is unimportant. The observation that inhibition of adenylyl cyclases suppressed [cAMP]_pm by ∼40% even in the presence of 1 mM IBMX indicates that the cAMP degradation in the sub-membrane space accounted for by IBMX-insensitive mechanisms is functionally important. Interestingly, it was recently demonstrated that the IBMX-insensitive PDE8B is expressed in islets and insulin-secreting cell lines, and that downregulation of the enzyme by RNA interference results in amplification of insulin secretion (Dov et al., 2008; Waddleton et al., 2008). We now found that PDE8B accounts for most, if not all, of the IBMX-insensitive PDE activity in the sub-membrane space. Accordingly, inhibition of adenylyl cyclases failed to counteract the IBMX-induced [cAMP]_pm elevation after knockdown of PDE8B. PDE8B is a cAMP-specific enzyme with a K_m of 0.15 µM (Soderling et al., 1998). Consistent with its high affinity for cAMP, this isoform contributed substantially to maintaining low basal levels of cAMP. Elevated basal [cAMP]_pm might explain why the magnitude of the glucose-induced [cAMP]_pm response appeared reduced in the knockdown cells. The amplitude of the IBMX responses were nevertheless elevated, which should not be surprising under conditions when most of the cAMP degrading capacity of the cell is inhibited. Furthermore, the oscillatory [cAMP]_pm signalling pattern was disturbed when expression of PDE8B was suppressed. The functional importance of this isoform was underlined by the finding that insulin pulsatility was abolished in glucose-stimulated knockdown cells. The loss of insulin pulsatility was associated with increased average secretion, which is in line with the previous observation that knockdown of PDE8B amplifies insulin secretion in rat islets (Dov et al., 2008).

cAMP oscillations have been observed in several systems and might be a widespread phenomenon fulfilling diverse functions in various cell types. In developing neurons, the motility response to axon guidance cues has been found to depend on cAMP oscillations (Nicol et al., 2007), and in insulin-secreting cells, [cAMP]_pm oscillations are translated to oscillations in insulin exocytosis (Dyachok et al., 2008). Moreover, the duration of a cAMP signal has been found to be crucial for the translocation of PKA catalytic subunits from the cytoplasm into the nucleus (Dyachok et al., 2006a; Dyachok et al., 2006b; Ni et al., 2011). cAMP oscillations could thereby provide a mechanism for spatially confining the activation of PKA-dependent effectors. The present study in insulin-secreting β-cells highlights an example where oscillations of cAMP in the sub-membrane space are generated by the periodic formation of the messenger by Adenylyl cyclases. Several PDE isoforms, including IBMX-sensitive PDE1 and PDE3 as well as the IBMX-insensitive PDE8B, contribute to shaping the [cAMP]_pm signal, but are not essential for generating the oscillations.

Analytical grade reagents and deionized water were used. Adrenaline, cilostamide, DDA, dipyridamole, dithiothreitol, EDTA, EGTA, HEPES, IBMX, MM-IBMX, 2-mercaptoethanol, papaverine, phenylmethanesulphonyl fluoride, poly-L-lysine, rolipram, sodium orthovanadate and Tris-HCl were purchased from Sigma (St Louis, MO, USA). Lipofectamine 2000, DMEM, trypsin, penicillin, streptomycin, glutamine and fetal calf serum were from Invitrogen (Carlsbad, CA, USA). Plasmid or adenoviral vectors encoding the two moieties of a cAMP translocation biosensor were used as previously described (Dyachok et al., 2006a; Dyachok et al., 2008). The sensor consists of a truncated and membrane-anchored PKA regulatory RIIβ subunit tagged with CFP and a PKA catalytic Cα subunit tagged with YFP. A plasmid encoding GRP1 (General receptor for phosphoinositides-1) fused to 4 tandem copies of (GFP₄–GRP1) was used to monitor plasma membrane phosphatidylinositol 3,4,5-trisphosphate [PtdIns(3,4,5)P₃] levels, which reflect insulin secretion with concomitant autocrine activation of insulin receptors and PI3-kinase (Dyachok et al., 2008). Lentiviruses encoding shRNA directed against PDE8B and a non-target control vector were purchased from Sigma. The PDE family-specific inhibitors were typically used at concentrations that were approximately ten times above their reported IC₅₀, but sometimes more than one concentration was tested.

Insulin-secreting MIN6 cells of passages 17–30 (Miyazaki et al., 1990) were cultured in DMEM containing 25 mM glucose and supplemented with 15% fetal calf serum, 2 mM glutamine, 70 µM 2-mercaptoethanol, 100 U/mL penicillin and 100 µg/mL streptomycin. Cells were seeded onto poly-L-lysine coated 25-mm coverslips and cultured to reach 50–60% confluence on the day of transfection. Transient transfection of the cAMP or PtdIns(3,4,5)P₃ biosensor plasmids was performed using Lipofectamine 2000 during 4 hours followed by culture in DMEM for 12–24 hours. Where indicated, MIN6 cells were treated with a mix of lentiviral vectors (at a multiplicity of infection of 10) expressing shRNA against PDE8B (5′-CCGGCCCAAACTTCATTTCCAGAAACTCGAGTTTCTGGAAATGAAGTTTGGGTTTTTG-3′, 5′-CCGGCCCATCACAAAGGTTATAAATCTCGAGATTTATAACCTTTGTGATGGGTTTTTG-3′, 5′-CCGGGCCATAGAAATAACAAGTGATCTCGAGATCACTTGTTATTTCTATGGCTTTTTG-3′) or the shRNA control vector (5′-CCGGCAACAAGATGAAGAGCACCAACTCGAGTTGGTGCTCTTCATCTTGTTGTTTTT-3′), 48 hours prior to the experiments. Islets of Langerhans were isolated from C57Bl6J female mice as described previously (Vieira et al., 2007). All procedures for animal handling and islet isolation were approved by the local animal ethics committee. After isolation, the islets were cultured for 1–4 days in RPMI-1640 medium containing 5.5 mM glucose, 10% fetal calf serum, 100 µg/ml penicillin and 100 µg/ml streptomycin at 37°C in an atmosphere of 5% CO₂ in humidified air. Data were obtained with cells from at least three independent islet isolations. The islets were infected with cAMP biosensor adenoviruses at a concentration of 10⁵ fluorescence forming units (FFU)/islet as described previously (Tian et al., 2011). Before the experiments, the cells or islets were transferred to a buffer containing 125 mM NaCl, 4.8 mM KCl, 1.3 mM CaCl₂, 1.2 MgCl₂ and 25 mM HEPES with a pH adjusted to 7.40 with NaOH, and pre-incubated for 30 minutes at 37°C in a humidified atmosphere with 5% CO₂. After the pre-incubation, the islets were applied onto poly-lysine-coated 25-mm coverslips where they immediately attached. β-cells were identified on the basis of their large size and negative response to adrenaline (Tian et al., 2011).

Total RNA was extracted from MIN6 cells using the RNEasy micro kit (Qiagen, Hilden, Germany). Real-time PCR was performed using Quanti Tect SYBR^(R) Green RT-PCR kit (Qiagen, Hilden, Germany), and the following primers were designed from the coding sequence of PDE8B: forward, 5′-GACTGATGAAGAGAAGAG-3′; reverse, 5′-ATGTCTGTTATGAAGTAGT-3′; PDE1C: forward, 5′-AAGCAGCAGAACGGTGACTT-3′; reverse, 5′-GGCAAGGTAATGCGACTTGT-3′; PDE3B: forward, 5′-CCAATTCCTGGCTTACCTCA-3′; reverse, 5′-GTGATCGTAATCGTGCATGG-3′; PDE4A: forward, 5′-CATCAATGTCCCACGATTTG-3′; reverse, 5′-TAAGTCCCGCTCCTGGAATA-3′; and β-actin: forward, 5′-GTTACAGGAAGTCCCTCACC-3′; reverse, 5′-GGAGACCAAAGCCTTCATAC-3′. PCR products were normalised to the housekeeping gene β-actin, and expression levels are given relative to the control according to the following formula: fold change = 2^ΔΔCt, where ΔΔCt = [Ct(PDE8B shRNA)-Ct(β-actin shRNA)Ct(PDE8B control)-Ct(β-actin control)].

Measurements of [cAMP]_pm or PtdIns(3,4,5)P₃ were performed as described previously with evanescent wave [total internal reflection fluorescence, (TIRF)] microscopy, using either a custom-built prism-based system (Idevall-Hagren et al., 2010) or an objective-based setup (Tian et al., 2011). The prism setup was built around an E600FN upright microscope (Nikon Corp. Tokyo, Japan). A helium-cadmium laser (Kimmon, Tokyo, Japan) provided 442 nm light for excitation of CFP, and the 514 nm line of an argon laser (ALC 60X, Creative Laser Production, Munich, Germany) was used to excite YFP. Interference filters (Semrock, Rochester, NY, USA) mounted in a filter wheel (Sutter Instruments, Novato, CA, USA) were used to select the appropriate wavelength. The merged laser beam was homogenised and expanded by a rotating light-shaping diffuser (Physical Optics Corp. Torrance, CA, USA) and refocused through a modified quartz dove prism (Axicon, Minsk, Belarus) with a 70° angle to achieve total internal reflection. The chamber was mounted on the custom-built stage of the microscope, such that the coverslip maintained contact with the dove prism by a layer of immersion oil. Fluorescence light was collected through a 40×, 0.8-NA water immersion objective (Nikon). The objective-based system consisted of an Eclipse Ti microscope (Nikon) with a TIRF illuminator (Nikon) and a 60×, 1.45-NA objective. The 458, 488 and 514 nm lines of an argon laser (ALC60X, Creative Laser Production) were used to excite CFP, GFP and YFP, respectively. The beam was coupled to the TIRF illuminator through an optical fibre (Oz Optics, Ottawa, Canada). In both evanescent wave microscope setups, fluorescence was detected with back-illuminated EMCCD cameras (DU-897, Andor Technology, Belfast, Northern Ireland) under MetaFluor (Molecular Devices Corp. Downington, PA) software control. Emission wavelengths were selected using filters [485 nm/25 nm half-bandwidth for CFP, 527/27 nm for GFP and 560/40 nm for YFP (Semrock Rochester, NY)] mounted in a filter wheel (Sutter Instruments). For time-lapse recordings, images or image pairs were acquired every 5 seconds. To minimize exposure of the cells to the potentially harmful laser light, the beam was blocked by a mechanical shutter (Sutter Instruments) between image captures.

Cells infected with control or PDE8B shRNA lentivirus were sonicated in a buffer containing 50 mM Tris pH 7.4, 2 mM EGTA, 1 mM EDTA, 250 mM sucrose, 1 mM dithiothreitol, 0.05 mM sodium orthovanadate, 1 mM phenylmethanesulphonyl fluoride, protease inhibitor cocktail (Sigma) and PhoSTOP (Roche). PDE activity was measured in duplicates as described previously (Murad et al., 1971). In order to determine non-PDE1/3/4 activity in the homogenates, assays were performed in the presence or absence of 50 µM MM-IBMX, 3 µM cilostamide and 10 µM rolipram. The effects of the non-selective PDE inhibitors, dipyramidole (100 µM) and IBMX (50 µM), were tested in the presence of the family-selective inhibitors.

Image analysis was performed using MetaFluor. The cAMP concentration was expressed as the ratio of CFP over YFP fluorescence after subtraction of the background. To compensate for variability in expression levels between different cells, the basal level ratio was normalised to 1. The GFP₄–GRP1 concentration in the plasma membrane was evaluated as the fluorescence intensity F in relation to the initial fluorescence intensity F₀ after subtraction of background (F/F₀). Time-average levels of [cAMP]_pm or PtdIns(3,4,5)P₃ were calculated by measuring the area under the curve followed by normalization for the elapsed time. All traces show original data, which has not been filtered or processed except for the traces in Fig. 1G,H, which have been corrected for base-line drift. Data are presented as means ± s.e.m. Statistical comparisons were assessed using the chi-square test or Student's t-test as appropriate.

We thank Heléne Dansk, Ing-Marie Mörsare and Ann-Kristin Holmén-Pålbrink for skilfull technical assistance. Author contributions were as follows: G.T. – experimental design, [cAMP]_pm measurements in mouse islets and shRNA-treated MIN6 cells, PtdIns(3,4,5)P₃ measurements, data analysis, manuscript preparation; J.S. – experimental design, [cAMP]_pm measurements in MIN6 cells, data analysis; Y.X. – experimental design, shRNA and PDE8B expression; H.S. – real-time PCR experiments; E.D. – experimental design, PDE assay experiments and data analysis; A.T. – conception of study, experimental design, data analysis and writing of manuscript. All authors read and approved the final version of the manuscript.