ADCY5 Couples Glucose to Insulin Secretion in Human Islets

Type 2 diabetes (T2D) is one of the foremost health challenges currently facing developed societies. This metabolic disease, which affects ∼8.3% of the adult population worldwide (1), usually reflects a failure of the β-cell mass to adapt output to increased peripheral insulin resistance. The resulting hyperglycemia and dyslipidemia lead to debilitating complications, ranging from kidney failure and blindness to cardiovascular disease and cancer (2). Although the maintenance of an adequate functional β-cell mass is critical for avoiding the development of diabetes (3), the molecular basis of β-cell failure is still poorly understood (4).

The mechanisms underlying glucose-stimulated insulin secretion from single β-cells involve uptake of the sugar via specific glucose transporters (5), enhanced ATP synthesis (6), and closure of ATP-sensitive K⁺ (K_ATP) channels (7). The consequent plasma membrane depolarization leads to Ca²⁺ influx (8) and exocytosis from secretory granules (9), both of which are further potentiated by “K_ATP-independent” amplifying signals (10). In addition, incretin hormones such as glucagon-like peptide 1 (GLP-1) and glucose-dependent insulinotropic peptide, released in response to food transit through the gut, potentiate insulin secretion in a glucose-dependent manner (11). Cognate receptor activation engages adenylate cyclases (ADCYs), enzymes that catalyze the generation of cAMP, a key intra- and intercellular signaling effector in the β-cell. Through its downstream interactions with protein kinase A and exchange protein activated by cAMP (Epac), cAMP drives changes including Ca²⁺ influx, intracellular Ca²⁺ mobilization (12), and the enhanced fusion competence of secretory granules (13). By contrast, glucose evokes more modest increases in intracellular cAMP (14), possibly via the stimulation of Ca²⁺-activated ADCYs such as ADCY1 and ADCY10 (15).

Both genetic (16) and environmental (4) risk factors conspire to determine the rate and extent of loss of insulin secretory capacity in T2D. Thus, the majority of genetic risk loci (∼70) currently identified by genome-wide association (GWAS) or familial studies alter functional β-cell mass while exerting little, or occasionally a beneficial, effect on insulin sensitivity (17). Of note, recent genetic studies have provided evidence that several pathways converging on β-cell cAMP signaling may influence T2D risk. For example, carriers of the major A-allele at rs11708067, or the C-allele at the neighboring single nucleotide polymorphism (SNP) rs2877716, lying on chromosome 3 in intron 3 of the ADCY5 gene, have an increased odds ratio of developing T2D (18). ADCY5 is a Ca²⁺-inhibited type III adenylate cyclase (19,20), and risk allele carriers present with elevated fasting glucose (21) but not impaired insulinogenic index or area under the curve (AUC)_{insulin/glucose} 2-h postoral glucose load (22–25). These data strongly imply that ADCY5 activity may be required for normal insulin release in response to glucose but not incretin, the latter largely accounting for the effects of oral glucose (26). Whether and how SNPs exert control over β-cell function by influencing ADCY5 expression remain, nonetheless, unclear.

Therefore, the aims of the current study were to 1) establish a link between genotype and ADCY5 mRNA levels, 2) silence ADCY5 expression in human islets using specific short hairpin (sh)RNAs, and 3) use in situ imaging approaches and hormone release assays to establish the role of the cyclase in regulating β-cell responsiveness to glucose and incretin.

Human islets were isolated from deceased heart-beating donors at transplantation facilities in Oxford, Geneva, Pisa, Edmonton, and Milan, with the relevant national and local ethics permissions, including consent from next of kin where required, and cultured as described (27). All studies involving human tissue were approved by the National Research Ethics Committee (NRES) London (Fulham), Research Ethics Committee no. 07/H0711/114.

Male and female C57BL/6 mice, 8–12 weeks of age, were killed by cervical dislocation and pancreatic islets isolated by collagenase digestion as previously described (28). Animal procedures were approved by the home office according to the Animals (Scientific Procedures) Act 1986 of the United Kingdom (PPL 70/7349).

DNA samples were genotyped for the SNP rs11708067 by restriction fragment length polymorphism analysis, as the G allele generated an HhaI site. A 250-bp region flanking the locus was amplified by PCR using Phire polymerase (Thermo) with the following primers: TCCGGAAGGCAAACACAGCA and AGCCAGGCTGCACCCAAGTG. The products were digested with HhaI and resolved by agarose gel electrophoresis.

Lentiviral particles carrying shRNA expression constructs against human ADCY5 were acquired from Sigma-Aldrich (Supplementary Table 1). Multiplicity of infection was calculated using Turbo-GFP particles on the same backbone (TRC 1.5; Sigma-Aldrich). Specificity of ADCY5 gene silencing was confirmed in both dissociated and intact islets, assuming 1,000–2,000 cells/islet for the latter. In all cases, lentiviral particles containing scrambled shRNA were used as controls (Con) and islets infected for 48–72 h.

cDNA encoding the cAMP sensor citrine/cerulean-Epac2-camps (29) was cloned into pShuttleCMV via HindIII and Xho1 sites before recombination with pAdEasy1 and virus production as described (6,30).

Relative mRNA abundance was quantified by quantitative RT-PCR (qRT-PCR) using SYBR green (31). Primers (Supplementary Table 2) were designed using PerlPrimer (32), specificity validated using a dissociation curve, and amplification efficiency was determined using a dilution series. Expression of each gene was normalized to cylcophilin A (Ppia) and N-fold change in mRNA expression versus Con calculated using the 2^−ΔΔCt.

ADCY5 mRNA levels were measured for expression quantitative trait locus analysis by qRT-PCR of RNA from isolated human islets as described earlier (33). The expression level of ADCY5 relative to TATA box–binding protein (TBP) was determined by qRT-PCR using Taqman primers and reagents (Supplementary Table 3) and the comparative Ct method (2^−ΔΔCT) used for subsequent calculations.

Islets were fixed overnight at 4°C in paraformaldehyde before application of primary antibodies against ADCY5/6 (cat. no. ab66037; Abcam) (34) and either guinea pig anti-insulin 1:200 or mouse anti-glucagon 1:1,000 (both DAKO). Revelation was performed with goat anti-rabbit Alexa-Fluor 488 and either goat anti–guinea pig Alexa-Fluor 568 or goat anti-mouse Alexa-Fluor 568 antibodies (both 1:500, Invitrogen). Images were acquired as previously described (27).

Insulin secretion was measured from groups of five to six islets per well, incubated for 30 min in 0.5 mL Krebs-HEPES-bicarbonate (KHB) solution (130 mmol/L NaCl, 3.6 mmol/L KCl, 1.5 mmol/L CaCl₂, 0.5 mmol/L MgSO₄, 0.5 mmol/L NaH₂PO₄, 2 mmol/L NaHCO₃, 10 mmol/L HEPES, and 0.1% [w/v] BSA, pH 7.4) at 37°C containing the indicated glucose and GLP-1 (7-36 human amide fragment) (Cambridge Bioscience) concentrations (35). Total islet proinsulin and insulin were measured after acid ethanol extraction and sonication using specific radioimmunoassays (EMD Millipore).

Islets were incubated with 3 μmol/L calcein-AM (Life Technologies) and 2.5 μmol/L propidium iodide (Sigma-Aldrich) before detection of absorbance/emission at 491/525 nm and 561/620 nm, respectively. The islet area occupied by dead cells was expressed as a unitary ratio versus that occupied by live cells. Apoptosis was assessed using a TUNEL staining kit (Promega) according to the manufacturer’s instructions, and islets were counterstained against insulin before detection of absorbance/emission at 491/525 nm and 561/620 nm, respectively.

Isolated islets were incubated (37°C, 95% O₂/5% CO₂) for 1 h in fluo2-AM (10 μmol/L) diluted in a bicarbonate buffer solution (120 mmol/L NaCl, 4.8 mmol/L KCl, 1.25 mmol/L NaH₂PO₄, 24 mmol/L NaHCO₃, 2.5 mmol/L CaCl₂, 1.2 mmol/L MgCl₂, and 3 mmol/L d-glucose; all Sigma-Aldrich). Functional multicellular Ca²⁺ imaging was performed using a Nipkow spinning-disk head and the resulting traces were normalized to minimum fluorescence (Fmin) (27). A near-identical distribution of Fluo-2 intensity was detected in Con and ADCY5 shRNA-treated islets under resting (3 mmol/L glucose) conditions, indicating that basal Ca²⁺ levels were likely unaffected by gene knockdown, assuming similar dye loading (Supplementary Fig. 1).

For imaging of cytosolic ATP/ADP with Perceval (6) or cAMP with Epac2-camps, islets were infected for 48 h with adenoviruses at multiplicity of infection 10–100, giving infection of the first 2–3 islet cell layers. Cells were imaged using a HEPES-bicarbonate buffer (120 mmol/L NaCl, 4.8 mmol/L KCl, 24 mmol/L NaHCO₃, 0.5 mmol/L Na₂HPO₄, 5 mmol/L HEPES, 2.5 mmol/L CaCl₂, and 1.2 mmol/L MgCl₂). For Perceval, absorbance/emission was 491/525 nm. For the Epac2-camps probe, excitation was delivered at 440 nm and emitted signals captured using cerulean (530 nm) and citrine (470 nm) filters. Förster resonance energy transfer (FRET) was calculated as the ratio of cerulean:citrine fluorescence and, for each experiment, expressed as a percentage of that obtained after maximal stimulation with 50 μmol/L forskolin (FSK). No differences in probe responses to FSK were detected in Con and ADCY5 shRNA-treated tissue (F/Fmin = 1.07 ± 0.01 vs. 1.06 ± 0.01 arbitrary units [AU], Con vs. ADCY5, respectively; P > 0.05).

ZIMIR, a membrane-bound probe that fluoresces upon binding of zinc (Zn²⁺) coreleased with insulin from granules, was used to dynamically monitor insulin secretion as previously described (27,36). Briefly, islets were incubated with 10 μmol/L ZIMIR for 2 h and imaged in bicarbonate buffer supplemented with 1 μmol/L EGTA. After acquisition (absorbance/emission = 491/525 nm), islets were divided into 20 subregions before extraction of intensity over time traces and analysis of amplitude and AUC.

Insulin-stained PFA-fixed islets were subjected to total internal reflection of fluorescence imaging to capture a superresolution snapshot of submembrane insulin granule distribution. The evanescent field was produced using a 561-nm laser (CrystaLaser) and a 100× 1.6 NA objective (Zeiss). Insulin granule density was expressed as a unitary ratio versus the total membrane area, the latter being clearly visible between adjacent cells.

Correlation and frequency analyses were performed as previously detailed (27,37). Briefly, intensity over time traces were extracted for each fluo-2–loaded cell using a region of interest before manual triage for glucose responsiveness based on rises above a 25% threshold. The Pearson product moment correlation was performed for all possible cell pair combinations and significance (P < 0.05) calculated versus the expected t distribution of independent R values. Functional connectivity maps showing the location of significantly correlated cell pairs were then constructed based upon correlation strength and position within the imaged field (x–y). Phase maps were compiled by converting the normalized intensity of each cell to a value between 1 and 100% and assigning this to a color using a light-dark ramp. Ca²⁺-spiking frequency was measured using the fast Fourier transform.

Data distribution was determined using D’Agostino omnibus or Shapiro-Wilk tests. Nonmultifactorial pairwise comparisons were made using Mann-Whitney U test or Student unpaired and paired t tests. Two-way ANOVA was used to assess interactions between multiple treatments (P < 0.01), and pairwise comparisons were performed using Bonferonni posttests. Expression quantitative trait locus data were stratified according to genotype, BMI, age, and sex. Factors included in the model were selected using the Akaike Information Criterion, resulting in the exclusion of age and BMI, as these were uninformative with regard to ADCY5 expression. To account for any prediction error, linear regression analyses were performed separately for both excluded variables, yielding nonsignificant relationships (P > 0.05). Owing to a significant interaction between sex and genotype in donors 22–70 years of age (P = 0.0476; two-way ANOVA), effects of genotype on mRNA abundance were assessed within sex group using one-way ANOVA. In all cases, analysis was performed using R (R Project), GraphPad Prism (GraphPad Software), IgorPro (Wavemetrics), and MATLAB (Mathworks), and results were considered significant at P < 0.05.

We sought firstly to characterize ADCY5 expression in tissue isolated from human and murine donors. Confirming previously published microarray data (33), SYBR Green qRT-PCR analyses revealed that ADCY5 and ADCY6 transcript levels were similar in human but not mouse islets (Fig. 1A). Immunohistochemistry using an antibody against ADCY5, and with some reported cross-reactivity against ADCY6, demonstrated the localization of both enzymes in the cytoplasm of insulin and glucagon immunopositive cells throughout individual human islets (Fig. 1B).

While GWAS provides statistically powerful information concerning associations between SNPs and T2D risk, it is unable to report on how the expression of genes at variant loci is altered in tissues implicated in glucose homeostasis. Since the elevated glucose levels associated with T2D may influence ADCY5 expression, we attempted to correlate genotype at SNP rs11708067 with ADCY5 mRNA abundance using tissue from a catalog of healthy donors. Allele frequency was close to that expected from previous studies (∼75% A) (18), and Taqman qRT-PCR analysis of RNA from isolated islets revealed approximately twofold lower mean ADCY5 expression in male AA versus AG carriers <70 years of age (Fig. 1C and Supplementary Table 4). When considered separately, neither age nor BMI significantly influenced ADCY5 expression for either genotype (Fig. 1D and E).

Given the above association between ADCY5 expression and fasting glucose levels and genotype in man, we next explored a role for this gene in human β-cell stimulus-secretion coupling. shRNAs directed against various sequences of the ADCY5 gene were delivered into dispersed cells or islets using replication-incompetent lentiviruses (n = 26 separate normoglycemic donors) (see Supplementary Table 5). Gene silencing efficiency was determined using qRT-PCR (Fig. 2A and B), and effects were confirmed at the protein level using immunohistochemistry and an antiserum raised against ADCY5/6 (Fig. 2C). Although mRNA levels were less affected by shRNA in intact than dissociated islets, this was most likely due to limited penetration of virus into the islet core (38); cells within the imaged layers were nevertheless typified by the near-complete absence of ADCY5 immunoreactivity. ADCY6 mRNA abundance was not altered by ADCY5 silencing, suggesting that the action of the shRNA on immunoreactivity was largely due to loss of the latter mRNA message (Fig. 2D). Cell viability and apoptosis indices were normal, excluding any effects of impaired cAMP signaling on β-cell survival (Fig. 2E and F).

The impact of ADCY5 silencing on glucose- or incretin-stimulated insulin secretory dynamics was subsequently assessed using Nipkow spinning-disk microscopy (27) to image Zn²⁺ coreleased from insulin-containing granules within individual ZIMIR-stained islets (27,36). The amplitude and AUC of glucose (11 mmol/L)-stimulated insulin release were markedly impaired in islets silenced for ADCY5 (Fig. 2G). By contrast, insulin-release dynamics in response to GLP-1 were subtly improved in tissue depleted for ADCY5, suggesting the presence of an intact incretin axis (Fig. 2H). The observations with ZIMIR were confirmed using conventional static incubation techniques followed by radioimmunoassay of supernatant, and we further detected no significant effect of gene silencing on KCl-stimulated insulin release (Fig. 2I–K). Donor variability was largely accounted for by a paired experimental design, and this was further supported by linear regression analyses, which revealed no relationship between age, BMI, and the magnitude suppression of glucose-stimulated insulin release in ADCY5-silenced islets (Supplementary Fig. 2). Insulin content was similar in Con and ADCY5 shRNA-treated islets, although there was a tendency toward an increased proinsulin-to-insulin ratio in the latter (Table 1), as expected from studies in patients harboring SNPs in or near the ADCY5 locus (25). Lastly, and in line with the secretory measures, total internal reflection of fluorescence imaging of insulin-stained islets confirmed that ADCY5 knockdown did not alter the distribution of granules located in the vicinity of the plasma membrane (Supplementary Fig. 3A and B).

Using Epac2-camps (29) to measure cAMP rises in β-cells (Supplementary Fig. 4), almost a threefold reduction in glucose- but not FSK-stimulated cAMP generation could be detected in ADCY5-silenced islets (Fig. 3A and B). Since cAMP elevations are linked to enhanced electrotonic coupling (39), the effects of ADCY5 depletion on the cell-cell communication processes underlying the intraislet regulation of insulin secretion were next investigated. Cytosolic free Ca²⁺ levels were thus used as a proxy to measure the electrical dynamics known to orchestrate Ca²⁺-dependent hormone secretion from the hundreds of cells residing within the first few microorgan layers; under the conditions used, Ca²⁺ changes chiefly reflect those within β-cells (27). In line with its dramatic effects on secretion, ADCY5 silencing suppressed both the amplitude and AUC of glucose-evoked Ca²⁺ rises throughout the imaged population (Fig. 3C–E and Supplementary Movies 1 and 2) (n = 3 donors; BMI range = 25.4–27.4; age range = 52–76 years). Furthermore, connectivity between individual β-cells, analyzed by large-scale mapping of long-term (∼30 min) evolutions in correlated cell-cell interactivity (40) and recently shown to be a key element in the insulin secretory response (27,41), was also reduced in ADCY5-silenced islets (Fig. 3F and G). Neither the number of responsive cells (Fig. 3H) nor the frequency of [Ca²⁺]_i oscillations (Fig. 3I) was affected by ADCY5 knockdown, suggesting that the absence of the cyclase was unlikely to impair insulin release by targeting the rhythmicity of a β-cell subpopulation. Implying that the effects of silencing were unlikely to reflect a long-term consequence of ADCY5 depletion, a similar degree of suppression of glucose-induced Ca²⁺ rises was obtained using 20 μmol/L NKY80, a relatively selective chemical inhibitor of the enzyme (IC₅₀ = 8.3 μmol/L, 132 μmol/L, and 1.7 mmol/L for ADCY5, 3, and 2, respectively) (42) (Supplementary Fig. 5A). Demonstrating the specificity of both shRNA and NKY80, the inhibitory actions of the drug were lost after ADCY5 silencing (Supplementary Fig. 5B).

Consistent with the lack of effect of gene silencing on GLP-1–induced insulin release, loss of ADCY5 failed to impact Ca²⁺ changes and even appeared to augment Ca²⁺ increases in response to the incretin without modulating the proportion of responsive cells (Fig. 4A and B and Supplementary Movies 3 and 4). Furthermore, GLP-1–mediated increases in β-cell–β-cell connectivity, critical for generating the acute (5–10 min) bursts in coordinated activity that underlie incretin potentiation of glucose-stimulated insulin secretion (27), were unchanged after treatment with anti-ADCY5 shRNA (Fig. 4C–E). In line with this, GLP-1R mRNA expression levels were unaffected by ADCY5 silencing (Fig. 4F).

We wondered whether the substantial reductions in glucose-induced Ca²⁺ rises and insulin release, detected in ADCY5-silenced islets, were accompanied by fulminant changes to β-cell glucose metabolism. To allow the real time recording of ATP dynamics specifically in β-cells, the expression of the recombinant probe Perceval was directed in human islets using an adenoviral vector (6,43). Confirming tropism of the virus for β-cells, as previously described in rodent islets (44), green fluorescent protein fluorescence was restricted to insulin-immunopositive cells (Fig. 5A). In response to increasing glucose concentrations (3–17 mmol/L), β-cells responded with large, coordinated, and nonoscillatory elevations in ATP-to-ADP ratio (Fig. 5B). Although significantly reduced, a glucose-induced increase in ATP-to-ADP ratio could still be detected after ADCY5 silencing (Fig. 5C). Further excluding a role for cAMP-independent effects on K_ATP or voltage-dependent Ca²⁺ channel (VDCC) activity, ADCY5 silencing did not alter Ca²⁺ rises induced by KCl applied in the absence and presence of the K_ATP channel opener, diazoxide (Supplementary Fig. 6A and B). Consistent with our previous findings in MIN6 β-cells (12), GLP-1 was able to provoke significant rises in ATP-to-ADP ratio under both low (3 mmol/L) and high (16.7 mmol/L) glucose conditions (Fig. 5D–F), as well as in the presence of ADCY5 silencing (Fig. 5G and H). Hence, an action of glucose to accelerate oxidative metabolism toward ATP synthesis is not a prerequisite for GLP-1 signaling.

To determine the relative contribution of metabolism to ADCY5-regulated β-cell function, ATP-to-ADP dynamics were imaged after exposure to increasing glucose concentrations. Whereas both Con and ADCY5-silenced islets responded normally to 5 and 8 mmol/L glucose (Fig. 6A), the latter failed to respond to further elevation of the sugar, and this deficit could be rescued using FSK to elevate cAMP (Fig. 6B). Cytosolic Ca²⁺ responses to 8 mmol/L glucose, however, remained suppressed after ADCY5 knockdown (Fig. 6C).

The aim of the current study was to explore the role of ADCY5 in the regulation of insulin secretion from human islets. We show that ADCY5 is required to link glucose-derived signals to the intracellular Ca²⁺ rises that principally drive insulin granule exocytosis. By contrast, incretins such as GLP-1 remain competent to evoke insulin release in the face of ADCY5 silencing (Fig. 7). Providing compelling evidence that defective ADCY5 action may contribute to impaired fasting glucose were the observations that mRNA expression was reduced in islets from subjects harboring risk alleles at rs11708067. Interestingly, ADCY5 mRNA levels were also reduced in β-cells from (ungenotyped) patients with T2D versus healthy donors (33), providing a further link with disease status.

Assessed by next-generation sequencing (RNA-Seq), ADCY5 mRNA occupies the top 13th centile of all mRNAs in human islets (45) and is the most strongly expressed member of the ADCY family in this tissue, with mRNA levels approximately twofold higher than those of ADCY1, the next most abundant isoform, and 50% higher than ADCY6 mRNA. Assuming similar levels of PPIA mRNA, our qRT-PCR results in isolated human islets are broadly consistent with this observation (Fig. 1A). Of note, however, analysis of laser-capture–dissected β-cells (33) (Gene Expression Omnibus public repository, accession no. GSE20966) and a β-cell–enriched fraction from human islets (46) revealed lower levels of ADCY5 mRNA (∼50th and ∼65th centiles, respectively), being in each case ∼50% of those of ADCY6. Intriguingly, in the current study, Adcy5 mRNA was barely detectable in mouse islets, being expressed at a level ∼40-fold lower than Adcy6 (Fig. 1). A similar, though less marked (approximately eightfold), preponderance of ADCY6 over ADCY5 mRNA also exists for rat islets (47), demonstrating marked species variability in the relative abundance of ADCY isoforms within this tissue. Since these variations do not appear to reflect the greater proportion of α-cells in human (48) versus rodent (49) islets, ADCY5 may serve a nonconserved function between mammalian species.

Levels of mRNA encoding ADCY8, previously reported to be regulated by glucose in rat and human islets (50), were recently shown to be much lower in human islets than either ADCY5 or -6 (45). Confirming these findings, ADCY8 mRNA was undetectable in four out of the five independent repeats that we conducted in human islets, and in the single experiment in which an amplicon was present, C_t values were incompatible with accurate quantification (37.05 ± 0.49 cycles).

Of other genes lying 250 kb on either side of rs2877716 and rs11708067, which might also conceivably mediate the observed effects on diabetes risk, MYLK transcripts are barely detectable in human islets (45,46). On the other hand, SEC22A, PDIA, and PTPLB are each detectably expressed and so may contribute. Nonetheless, our observations on the effects of ADCY5 silencing are entirely consistent with the observed phenotype of subjects possessing risk alleles (22).

The strong dependence of glucose-stimulated insulin secretion upon ADCY5 expression in human islets is surprising given the abundant presence of ADCY6 and the redundancies inherent to the cAMP signaling cassette (51). Although the above differences at the mRNA level are not necessarily reflective of protein quantity or enzymatic activity, the observed sensitivity to ADCY5 depletion may reflect either microcompartmentalization of ADCY5-containing complexes in proximity to Ca²⁺ influx and release sites (52) or a requirement for ADCY5 in cAMP-independent signaling processes. Thus, while our data indicate that ADCY5 couples glucose to insulin secretion by translating a glucose signal into cAMP generation (53), GLP-1 receptors may engage alternative ADCY family members to modify cAMP dynamics. Further studies will be required to explore the latter possibility (Fig. 7). Of note, global deletion of ADCY5 in mice leads to longevity and the resistance of cardiomyocytes to oxidative stress via the upregulation of Ras-MAPK signaling (20). A similar mechanism might therefore contribute to the small but significant enhancement of GLP-1 signaling observed here in human β-cells after ADCY5 silencing.

Although ADCY5 is Ca²⁺ inhibited in vitro (19), the range of Ca²⁺ concentrations over which inhibition occurs (>10 μmol/L) comfortably exceeds normal intracellular concentrations of these ions in β-cells (8). Whether ADCY5 is therefore a direct target for activation by intracellular signals generated by glucose (ATP, etc.) or is, rather, a passive but essential element of a glucose-activated signaling pathway leading to Ca²⁺ influx and insulin release remains to be established.

Recent studies have shown that the intraislet regulation of cell-cell communication is critical for the proper generation of hormone release after secretagogue challenge (27). Providing further evidence that the islet context is an important player in insulin secretion was the observation in the current study that β-cell responses to glucose were less coordinated in islets depleted of ADCY5. While the mechanisms underlying this phenomenon remain obscure, cAMP has been shown to alter β-cell gap junction conductance (39), and the resultant enhanced intercellular coupling may contribute to the insulin-raising actions of cAMP-elevating agents such as glucose and incretins. Whereas we recently showed that knockdown of connexin 36 markedly reduced coordinated cell responses to incretin, more modest effects on glucose action were observed (27). The impact of cAMP on glucose-induced cell connectivity may therefore also stem from perturbed paracrine signaling circuits between β-, α-, and other cell types, in addition to enhanced electrotonic coupling. Likewise, an intriguing possibility is that ADCY5 suppression may also affect glucagon and, potentially, GLP-1 (54) secretion from neighboring α-cells to affect β-cell “glucose competence.” In any case, ADCY5-depleted islets still displayed impaired Ca²⁺ responses even when metabolic dysfunction was accounted for (Fig. 6B), supporting the view that other downstream processes, including cell-cell communication, are targeted by cAMP signaling.

Remarkably, GLP-1 evoked Ca²⁺ rises and hormone release even when glucose-stimulated insulin secretion was abolished by ADCY5 silencing. This implies that GLP-1 signaling—generally believed to hinge on increased intracellular cAMP levels (55)—is sufficient under these conditions to elicit closure of K_ATP and induce Ca²⁺ influx and exocytosis. Indeed, studies by Miki et al. (56) have demonstrated that GLP-1 is still able to evoke insulin secretion in the complete absence of K_ATP channels. Thus, the glucose dependency of GLP-1–induced insulin secretion is unlikely solely to reflect the effects of the sugar on oxidative metabolism and, hence, the closure of K_ATP channels, especially in light of data showing the absence of GLP-1–induced Ca²⁺ rises in human islets incubated at nonpermissive glucose concentrations (27). Rather than a simple summation of glucose and GLP-1–derived Ca²⁺ signals, the ATP/ADP responses to the latter, readily detectable at low glucose concentrations, may instead be converted into a depolarizing stimulus through a complex interplay between K_ATP-dependent and -independent pathways (10).

In summary, we describe here a novel role for the GWAS-identified gene ADCY5 in the normal regulation of insulin secretion from human islets of Langerhans. Thus, we elucidate a pathway that converts ADCY5 gene polymorphisms into defective β-cell function. Since β-cell decompensation is a hallmark of T2D pathogenesis irrespective of genotype, ADCY5 may provide a useful target for the restoration of insulin release in man.

Acknowledgments. The authors thank Drs. Tatsuya Kin and Doug O'Gorman from the Clinical Islet Isolation Laboratory in Edmonton for providing human islets for this study.

Funding. These studies were supported by a Diabetes UK R.D. Lawrence Research Fellowship (12/0004431) to D.J.H. and by a Wellcome Trust Senior Investigator award (WT098424AIA), a Medical Research Council Programme award (MR/J0003042/1), a Diabetes UK Project grant (11/0004210), and a Royal Society Wolfson Research Merit award to G.A.R. Isolation of human islets was supported by JDRF awards (31-2008-416) to D.B. and (31-2008-413) to L.P. (European Consortium for Islet Transplantation). P.R.J. and S.J.H. were supported by grants from the National Institute for Health Research Biomedical Research Centre, Oxford. W.-H.L. was supported by grants from the JDRF (37-2011-21) and National Institutes of Health (R01-GM-077593). The work leading to this publication has received support from the Innovative Medicines Initiative Joint Undertaking under grant 155005, resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013), and European Federation of Pharmaceutical Industries and Associations companies’ in-kind contribution to G.A.R.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. D.J.H. supervised the research, conceived and designed the study, performed the experiments, performed analysis, and wrote the manuscript with input from all authors. R.K.M. performed the experiments and analysis and conceived and designed the study. L.M. and T.J.P. performed the experiments and analysis. S.G.B. and F.S. performed the experiments. K.L.E., D.M.F.C., D.L., and W.-H.L. provided reagents. M.B., P.M., V.L., D.B., L.P., P.R.J., S.J.H., and A.M.J.S. isolated and provided human islets. G.A.R. jointly supervised the research, conceived and designed the study, performed analysis, and wrote the manuscript with input from all the authors. G.A.R. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 73rd Scientific Sessions of the American Diabetes Association, Chicago, IL, 21–25 June 2013.