Voltage-Gated Ion Channels in Human Pancreatic β-Cells: Electrophysiological Characterization and Role in Insulin Secretion

With appropriate ethical approval and clinical consent, pancreatic islets were isolated in the Diabetes Research and Wellness Foundation Human Islet Isolation Facility from human pancreata retrieved from nondiabetic, heart-beating donors. This study is based on 34 islet preparations. For patch-clamp experiments, islets were dispersed into single cells immediately after preparation by incubation in Ca²⁺-free buffer followed by trituration. The cells were cultured in RPMI-1640 medium containing 10 mmol/l glucose and 2 mmol/l l-glutamine.

ω-Conotoxin GVIA, SNX482, ω-agatoxin IVA, and stromatoxin-1 were purchased from Alomone Labs (Jerusalem, Israel) or the Peptide Institute (Osaka, Japan). Iberiotoxin (IbTX) was from Bachem (St. Helens, U.K.). All other chemicals were purchased from Sigma-Aldrich.

Insulin secretion was measured as described elsewhere (15). Briefly, batches of 10–20 islets (in triplicates) were preincubated in 1 ml Krebs-Ringer buffer supplemented with 1 mmol/l glucose for 1 h followed by a 1-h incubation in 1 ml of test Krebs-Ringer buffer medium supplemented as indicated.

Patch-clamp experiments were performed using an EPC-9 amplifier and Pulse software (HEKA, Lamprecht, Germany). All electrophysiological experiments were performed at 32–33°C using the standard or perforated-patch whole-cell configurations.

K⁺ currents were recorded in an extracellular solution composed of (mmol/l) 138 NaCl, 5.6 KCl, 2.6 CaCl₂, 1.2 MgCl₂, 5 HEPES, and 5 glucose (pH 7.4, adjusted with NaOH). For recording Na⁺ and Ca²⁺ currents, 10 mmol/l tetraethylammonium (TEA) was added and NaCl correspondingly reduced. For Ca²⁺ current measurements, 0.1 μg/ml tetrodotoxin was added additionally. For Na⁺ current recordings, extracellular CaCl₂ was replaced equimolarly with MgCl₂ and 1 mmol/l CoCl₂ included in the medium.

The intracellular solution for K⁺ current measurements contained (mmol/l) 120 KCl, 1 MgCl₂, 10 EGTA, 1 CaCl₂, 10 HEPES, and 3 MgATP (pH 7.2, KOH). Na⁺ and Ca²⁺ current measurements were made after equimolar substitution of KCl by CsCl. For capacitance measurements, the pipette solution contained (mmol/l) 125 Cs-glutamate, 10 CsCl, 10 NaCl, 1 MgCl₂, 5 HEPES, 0.05 EGTA, 3 MgATP, and 0.1 cAMP (pH 7.2, CsOH). Glucose- and tolbutamide-induced electrical activity was recorded in the perforated-patch whole-cell configuration as previously reported (16). Biocytin (0.5 mg/ml; Invitrogen) was added to all pipette solutions to facilitate subsequent establishment of cell identity.

Immunolabeling with antibodies against insulin, glucagon, and somatostatin and detection of biocytin-loaded cells was performed essentially as described previously (17).

Gene expression profiling of ion-channel isoforms was performed by RT-qPCR on human islet total RNA (for details see the online appendix [available at http://dx.doi.org/10.2337/db07-0991]).

All data points are expressed as means ± SE of indicated number of experiments. All experiments were performed using islets/cells from at least two different donors. Inhibitory effects of channel blockers on glucose-induced insulin secretion are given after subtraction of basal release. Statistical significance was evaluated using Student's t test. A more detailed description of the experimental procedures is provided in the online appendix.

The identity of all β-cells used for electrophysiological analysis in this study was confirmed by immunocytochemistry. Human β-cells thus identified had a membrane capacitance (C_slow) of 9.9 ± 0.3 pF (n = 207) (online appendix Fig. 1).

Voltage-gated K⁺ currents were elicited by 500 ms voltage-clamp depolarizations from a holding potential of −70 mV. Outward K⁺ currents became detectable during depolarizations to −30 mV and beyond (Fig. 1A). The activation kinetics of the K⁺ current elicited by depolarization to +10 mV was estimated using a Hodgkin-Huxley m⁴ model. The time constant (τ) of activation averaged 1.9 ± 0.1 ms (n = 31). In 26 of 31 cells, current inactivation was best described as the sum of two exponentials (Fig. 1B). The average time constants (τ) were 41 ± 7 ms for the fast (transient) component and 1.8 ± 0.1 s for the slowly inactivating (sustained) current. The transient component accounted for 28 ± 2% of the total current amplitude at +10 mV. In the remaining five cells, the inactivation could be well described using a single time constant of 1.8 ± 0.3 s. In cells with a transient current component, a small “shoulder” in the current-voltage relationship (I-V) relationship was apparent at voltages between +40 and +60 mV (Fig. 1A and C).

The peak outward current was sensitive to TEA; 1 and 10 mmol/l of this broad-spectrum K⁺ channel blocker inhibited 50 ± 3% (n = 9) and 83 ± 3% (n = 14), respectively (Fig. 1D). The Kv2.1/2.2 antagonist stromatoxin-1 selectively inhibited the slowly inactivating outward current (Fig. 1E). In a series of eight experiments, the sustained current component (measured at the end of a 250-ms depolarization to +20 mV) was reduced by 45 ± 6% (P < 0.001), while the peak current decreased only by 15 ± 4% (P < 0.01).

The Ca²⁺ channel blocker Co²⁺ selectively abolished the transient current (Fig. 2A) and removed the shoulder in the I-V relationship (Fig. 2B). IbTX, a specific blocker of large-conductance Ca²⁺-activated K⁺ channels (BK channels), also selectively blocked the transient K⁺ current (Fig. 2C), and its effect was indistinguishable from that of Co²⁺ (compare Fig. 2A and C). The IbTX-sensitive current was reconstructed by subtracting the current obtained in the presence of the blocker from that measured under control conditions. The isolated IbTX-sensitive current exhibited a bell-shaped voltage dependence with a peak of 41 ± 11 pA/pF (n = 9) at +30 mV (Fig. 2D), comparable with the 52 ± 23 pA/pF (n = 3) of the Co²⁺-sensitive current (not shown).

Figure 2E (inset) shows the total and IbTX-resistant currents evoked by depolarization to −20 mV. Whereas the current before IbTX activated rapidly, the IbTX-insensitive current developed more slowly. Figure 2E summarizes the relationship between voltage and time constants of activation (τ_m) of IbTX-sensitive and -resistant currents. The IbTX-sensitive and -insensitive currents elicited by depolarizations to +10 mV inactivated with time constants of 22 ± 6 ms (n = 7) and 1.6 ± 0.3 s (n = 7), respectively (not shown). These values are comparable with the time constants for the fast and slowly inactivating components under control conditions described above. The magnitude of the IbTX-sensitive component at +10 mV was ∼50% of that of the IbTX-insensitive component (approximately one-third of the total), in good agreement with the fraction of the current showing rapid inactivation under control conditions (28%, see above).

The effects of the K⁺ channel blockers on β-cell electrical activity are shown in Fig. 3. TEA (10–20 mmol/l) increased the peak voltage of the action potentials induced by 6 mmol/l glucose (n = 3) (Fig. 3A) or tolbutamide (100 μmol/l; n = 4) from −13 ± 5 mV to 4 ± 7 mV. The K_V2.1/2.2 blocker stromatoxin had weak effects on electrical activity; a significant increase in spike height was observed in only one of seven cells (Fig. 3B) and the action potentials peaked at −21 ± 3 mV in both the absence and presence of the blocker (n = 4 for glucose, n = 3 for tolbutamide). Addition of IbTX significantly increased the amplitude of glucose-induced action potentials in two of five cells (Fig. 3C), whereas action potential firing was inhibited in the remaining three cells. The latter effect was accompanied by hyperpolarization and an increase in resting membrane conductance (n = 2) (not shown), presumably reflecting activation of K_ATP channels. When IbTX was tested in the presence of tolbutamide (100 μmol/l), it consistently increased the amplitude of action potentials (n = 3). Similar effects of IbTX were observed in current-injection experiments (online appendix Fig. 2A).

Glucose at 6 mmol/l (a concentration attained postprandially in nondiabetic individuals [18]) and 20 mmol/l stimulated insulin secretion 3.7- and 7.7-fold over basal (1 mmol/l), respectively (Fig. 3D). TEA enhanced insulin secretion at 6 and 20 mmol/l glucose by 85 and 94%, respectively (Fig. 3D). By contrast, stromatoxin had no significant effects on glucose-induced insulin secretion (Fig. 3E), even when tested at a concentration of 1 μmol/l (n = 3 from one donor) (data not shown) (19). We ascertained that stromatoxin remained a blocker of the delayed outward current under the experimental conditions used in the secretion assays (i.e., presence of BSA) (data not shown). Whereas IbTX was without effect at 20 mmol/l glucose (data not shown), it abolished the stimulatory effect of 6 mmol/l glucose (Fig. 3F). When tested in the simultaneous presence of 6 mmol/l glucose and 0.1 mmol/l tolbutamide, however, IbTx stimulated insulin secretion by 73% (Fig. 3G).

When the pipettes were filled with a Cs⁺-containing medium, voltage-clamp depolarizations to 0 mV elicited inward currents consisting of a very transient (∼2 ms) and a more slowly inactivating component (Fig. 4A). Whereas the sustained component was suppressed by removal of extracellular Ca²⁺ and addition of the broad-spectrum Ca²⁺ channel blocker Co²⁺, the transient current was abolished by the selective Na⁺ channel blocker tetrodotoxin (TTX). Voltage-gated Na⁺ currents were activated by depolarizations to −30 mV and above and reached a maximal amplitude of 40 ± 7 pA/pF at 0 mV (n = 8) (Fig. 4B and C). The voltage dependence of inactivation was characterized using a two-pulse protocol (Fig. 4D). Inactivation was observed after prepulses positive to −80 mV. A Boltzmann fit yielded values for voltage of half-maximal inactivation (V_h) and slope factor (n_h) of −42 ± 2 mV and 6 ± 1 mV (n = 9), respectively (Fig. 4E).

We studied the effect of TTX on glucose-induced electrical activity (Fig. 4F). In the presence of 6 mmol/l extracellular glucose, the cell-generated action potentials originating from ∼−60 mV and reaching up to 0 mV. Addition of TTX reversibly reduced the peak voltage of the action potentials from −12 ± 2 mV to −24 ± 3 mV (n = 2). Similar effects were observed in current injection experiments (n = 11) (online appendix Fig. 2B). As shown in Fig. 4G, the TTX-sensitive Na⁺ current is important for glucose-induced insulin secretion, and secretion elicited by 6 and 20 mmol/l glucose was reduced by 70 and 55%, respectively, in the presence of TTX.

Ca²⁺ currents became detectable during depolarizations to −50 mV and peaked at 0 mV (Fig. 5A), where the current density was 14 ± 1 pA/pF (n = 41). The activation kinetics of the Ca²⁺ current was described using a Hodgkin-Huxley m² model. At 0 mV, the current activated with a time constant of 0.41 ± 0.02 ms (n = 28). The inactivation of the current was biphasic in most cells. The time constants for the rapid (comprising 35 ± 3%) and slow components (43 ± 3%) averaged 6.8 ± 0.4 ms (n = 25) and 65 ± 15 ms (n = 28), respectively.

The Ca²⁺ channel subtypes underlying the voltage-gated Ca²⁺ current were established using specific blockers. A representative recording is shown in Fig. 5B. In this experiment, the current was reduced by >90% by sequential application of first the L-type blocker isradipine alone, followed by a combination of isradipine and the P/Q-type inhibitor ω-agatoxin IVA and finally isradipine together with the T-type antagonist NNC 55-0396 (see also online appendix Fig. 3). Table 1 summarizes the inhibitory effects of the Ca²⁺ channel blockers on the peak and integrated Ca²⁺ currents evoked by depolarizations from −70 to 0 mV. The cumulative inhibitory effects of isradipine, ω-agatoxin IVA, and NNC 55-0396 amounted to 91%. The N-type blocker ω-conotoxin GVIA and the R-type blocker SNX-482 exerted negligible effects.

The voltage-dependent inactivation of the T-type Ca²⁺ current was studied using a two-pulse protocol (Fig. 5C). The experiments were performed in the presence of isradipine and (in some cells) ω-agatoxin IVA to block L- and P/Q-type Ca²⁺ currents. A Boltzmann fit to the data points yielded values for V_h and n_h of −64 ± 2 mV and 8 ± 1 mV (n = 13), respectively.

The currents flowing through the different channel subtypes were isolated by subtracting the responses recorded after application of the respective blockers from those obtained before blocker application (Fig. 5D). I-V relationships for the total Ca²⁺ current and the individual components are shown in Fig. 5E. T-type currents activated at −50 mV and reached a peak between −40 and −30 mV. L-type currents were first seen during depolarizations to −40 mV and reached a maximum between −20 and −10 mV. P/Q-type currents became significant only at potentials beyond −20 mV and peaked at 0 mV. The effect of ω-agatoxin IVA showed a slow onset and was maximal only 3–4 min after addition (not shown) (20). Time-dependent rundown of T-type Ca²⁺ channels might therefore account for the apparent ω-agatoxin sensitivity of a current component at voltages more negative than −20 mV.

In insulin release experiments, isradipine inhibited glucose-stimulated (20 mmol/l) insulin release and reduced secretion below baseline levels (Fig. 6A). By contrast, both SNX-482 and ω-conotoxin GVIA were without inhibitory effect, whereas ω-agatoxin IVA exerted a moderate inhibitory effect (−31%) (Fig. 6B). The T-type channel blocker NNC 55-0396 was without effect on insulin secretion at 20 mmol/l glucose (not shown). When insulin secretion was evoked by 6 mmol/l glucose, addition of NNC 55-0396, ω-agatoxin IVA, and isradipine reduced glucose-stimulated release by 59, 71, and 100%, respectively (Fig. 6C and D), whereas ω-conotoxin remained ineffective (not shown).

Figure 6E shows a cell electrically active at 20 mmol/l glucose. In all cells tested, isradipine completely suppressed electrical activity induced by glucose (6 or 20 mmol/l) (n = 3) or tolbutamide (n = 2). The T-type antagonist NNC 55-0396 also reduced action potential amplitude and frequency (Fig. 6F). The peak voltage attained during the action potential was reduced from −28 ± 2 mV under control conditions to −33 ± 3 mV (P < 0.05) (n = 3) in the presence of NNC 55-0396. Data obtained in membrane potential recordings echo those in the current-injection experiments (online appendix Fig. 2C and D).

Capacitance measurements were used to investigate depolarization-evoked exocytosis in human β-cells. While 20-ms depolarizations usually failed to evoke a clear (>10 fF) response, longer depolarization pulses triggered progressively larger capacitance increases (Fig. 7A). In many β-cells, exocytosis continued beyond the depolarization. Figure 7B plots the increase evoked by the depolarization against pulse duration. The average response during 500-ms depolarizations was 41 ± 5 fF/pF.

Figure 7C shows the relationship between the integrated Ca²⁺ current and exocytosis. Exocytosis was small for integrated Ca²⁺ currents with a charge of <0.4 pC/pF but then increased supralinearly with charge entry. The relationship was well described by a fourth-order polynomial function, compatible with cooperative binding of Ca²⁺ to the Ca²⁺ sensor of exocytosis (21).

We studied the voltage dependence of exocytosis by applying depolarizing pulses from −70 mV to voltages between −40 and +40 mV (Fig. 7D). Depolarizations to −20 mV or below were largely ineffective, but the responses then increased steeply, with a maximum being attained at 0 mV and a secondary reduction at more positive voltages. The observed voltage dependence of exocytosis most closely resembles that of the P/Q-type Ca²⁺ channels (cf. Figure 5E).

We studied the effects of Ca²⁺ channel blockers on depolarization-evoked exocytosis. Whereas ω-agatoxin exerted a strong inhibitory action on exocytosis evoked by a 500-ms depolarization to 0 mV, the effect of isradipine was weak (Fig. 7E). On average, isradipine and ω-agatoxin reduced exocytosis by 31 ± 15% (NS) (n = 6) and 80 ± 3% (P < 0.001) (n = 5), respectively.

We quantified mRNA expression of voltage-gated ion channels in nine human islet preparations (online appendix Table 1). All known isoforms of Na⁺ channels were screened. Of the α-subunits, Na_V1.7 and Na_V1.6 dominated, while type 1β constitutes 80% of the total β-subunit expression. Of the Ca²⁺ channels, Ca_V3.2 (α_1G; T type), Ca_V1.3 (α_1D; L type), and Ca_V2.1 (α_1A; P/Q type) isoforms were most abundant and account for 48, 28, and 12% of the transcript numbers, respectively. A selection of voltage-gated K⁺ channels based on Yan et al. (22) was screened. The expression of K_V channels was dominated by K_V2.1 and K_V2.2 (24 and 73%, respectively). The α-subunit of the large conductance Ca²⁺-activated K⁺ channels (BK channels) was highly expressed, and of the β-subunits β₂ predominated (80%).

We have characterized the voltage-gated membrane currents in identified human β-cells, their molecular composition, and their involvement in exocytosis, electrical activity, and glucose-induced insulin secretion. This was facilitated by the access to novel and selective blockers. Based on these data, we provide a model that outlines the respective roles of the different voltage-gated ion channels in stimulus secretion coupling in human β-cells.

The voltage-gated K⁺ current in human β-cells consists of at least two different components. A transient component activates rapidly upon membrane depolarization, is dependent on Ca²⁺ influx, and is blocked by IbTX. These properties suggest that it is carried by large-conductance Ca²⁺-activated K⁺ channels (BK channels). BK currents exhibited a typical bell-shaped voltage dependence, with decreasing amplitude at potentials >30 mV due to reduced Ca²⁺ entry. It can be noted that the peak BK current is observed at a potential ∼30 mV more positive than the peak Ca²⁺ current (compare Figs. 2D and 5E). This rightward shift reflects both the intrinsic voltage dependence of the channels, with (at fixed Ca²⁺ levels) increased open probability at depolarized potentials (23), and the increasing driving force for K⁺ entry. Similar to studies in chromaffin cells (24) and mouse β-cells (25), BK currents were recorded despite the presence of a high concentration of EGTA in the intracellular solution. This suggests that BK channels and voltage-gated Ca²⁺ channels are closely colocalized (26). In mouse β-cells, BK channels do not play a major role in glucose-induced electrical activity or insulin secretion (27). By contrast, in human β-cells blockade of BK channels increases spike amplitude. In some cells, addition of IbTX suppressed action potential firing. This effect appeared to result from an unexpected ability of IbTX to activate K_ATP channels, which also accounts for the inhibition of glucose-induced insulin secretion. When insulin secretion was instead measured in the presence of tolbutamide, IbTX, as expected from the electrophysiological data, stimulated insulin secretion.

The second component of the K⁺ current developed with a delayed time course. This component was unaffected by Co²⁺ and IbTX but inhibited by stromatoxin and TEA; inhibition by the latter compound was half maximal at ∼1 mmol/l. The current is likely to be due to delayed rectifying K⁺ channels. K_V2.1 channels, which are dominant in rodent cells (28), are also expressed in human β-cells (22). However, at the mRNA level K_V2.2 channels are more abundant in human islets. In rodent cells, blockade of K_V2.1/2.2 channels stimulates insulin secretion (19,28). By contrast, in human islets the K_V2.1/2.2 blocker stromatoxin had no major effect on insulin release and electrical activity. It has been suggested that human β-cells also express A-type K⁺ currents (29), but we have so far detected such currents only in non–β-cells in human islets (not shown). Collectively, these data suggest that BK channels are particularly important for spike repolarization. Their activation kinetics and voltage dependence make them ideally suited for this task.

In agreement with previous studies in human β-cells (10,12), but in marked contrast to mouse β-cells (30–32), the voltage-gated Na⁺ current could be activated from physiologically relevant membrane potentials (−70 mV and more positive). Using TTX, we could demonstrate the significance of these channels for action potential generation and insulin secretion. At the mRNA level, human islets express approximately equal amounts of Na_v1.6 and Na_v1.7. The latter channels are involved in nociception (33), and Na_v1.7-specific antagonists are considered as analgesics. The possibility that such channels are involved in insulin secretion suggests that the use of such drugs may cause impaired insulin secretion as a side effect.

In agreement with previous studies (5,11), we show that L-type Ca²⁺ channels are expressed in human β-cells. Blockade of L-type Ca²⁺ channels using the selective antagonist isradipine leads to complete inhibition of glucose-induced insulin secretion. Our data suggest that the importance of L-type channels predominantly results from their essential role in the generation of electrical activity, whereas their role in exocytosis is modest (∼30%). The significance of L-type channels for the generation of electrical activity may reflect their voltage dependence: activation commences at voltages as negative as −40 mV (Fig. 5E). The PCR data suggest that the L-type Ca²⁺ channel is of the Ca_V1.3 (α1D) subtype. The fact that the isradipine-sensitive current activates at more negative voltages in human than in mouse β-cells is consistent with this observation (34).

Although expression of P/Q-type Ca²⁺ channels in human β-cells has been reported earlier (8), their relative contribution to the total Ca²⁺ current and their function in stimulus secretion coupling have been unclear. We now show that P/Q-type Ca²⁺ channels account for ∼45% of the integrated whole-cell Ca²⁺ current and that they play a critical role in depolarization-evoked exocytosis and glucose-induced insulin secretion, especially at low glucose concentrations (6 mmol/l). At variance with mouse β-cells (35), we found no electrophysiological evidence for the presence of R-type Ca²⁺ channels. In fact, they are not at all expressed in human islets (online appendix Table 1).

We further confirm that T-type Ca²⁺ channels are also expressed (5) and contribute to electrical activity (10) in human β-cells. Using a recently developed, more selective antagonist (NNC55-9036) (see online appendix Fig. 3), we obtained evidence that T-type channels are involved in insulin release evoked by 6 mmol/l glucose but not by 20 mmol/l glucose. We acknowledge that NNC55-9036 is not ideal in inhibiting non–T-type Ca²⁺ currents also. However, the nonspecific effects (unlike those on the T type) were reversible. For secretion studies, islets were therefore preincubated for 15 min with the blocker, followed by a 10-min washout phase. In patch-clamp experiments using the same protocol, only the peak (T-type) Ca²⁺ current was significantly reduced after pretreatment with NNC55-9036, whereas the sustained current (non–T type) was unaffected (see legend to online appendix Fig. 3). PCR analysis suggests that the T-type Ca²⁺ current in human β-cells is of the Ca_V3.2 subtype.

Based on the findings in the present and previous studies, we propose that at physiological glucose concentrations (∼6 mmol/l) the closure of K_ATP channel depolarizes the β-cell membrane to potentials above −55 mV. The activation of T-type (at voltages above −60 mV) and L-type Ca²⁺ channels (above −50 mV) (Fig. 5E) initiates the action potential. During the upstroke of the action potential, voltage-gated Na⁺ channels also open (above −40 mV) (Fig. 4B and C), leading to a further acceleration of the upstroke and sufficient depolarization to activate P/Q-type Ca²⁺ channels (above −20 mV). Ca²⁺ influx via P/Q-type (and to a lesser extent L-type) Ca²⁺ channels directly triggers exocytosis of insulin granules. The β-cell is repolarized by the activation of Ca²⁺-activated BK channels, with K_V2.1/2.2 channels playing only a minor role. The scenario outlined above is consistent with the observation that the inhibitory actions of TTX (see also 10), NNC55-0936, and ω-agatoxin are weaker at 20 than at 6 mmol/l glucose. As shown in Figs. 4F and 6F, inhibition of T-type Ca²⁺ channels and Na⁺ channels reduces the action potential amplitude. A reduction of spike height by 15–20 mV (from the normal peak voltage of −10 to 0 mV) will result in >50% reduction of both the P/Q-type Ca²⁺ current (Fig. 5E) and exocytosis (Fig. 7D). Our data suggest that P/Q-type Ca²⁺ channels are more tightly coupled to exocytosis than L-type Ca²⁺ channels. Indeed, the exocytotic responses were small during depolarizations to −20 mV, a voltage associated with the maximum activation of the L-type Ca²⁺ channels (Fig. 5E).

It is evident that human and mouse β-cells differ in many respects. Thus, some channels which are not considered functionally important in mouse β-cells (like the BK channels, T- and P/Q-type Ca²⁺ channels, and voltage-gated Na⁺ channels) play critical roles in human β-cells. Conversely, R-type Ca²⁺ channels and K_V2.1 channels appear less important in human cells than suggested by previous work in mice. As discussed above, this may have an impact on drug development. It seems unlikely that the differences between human and rodent β-cells are confined to ion channels. Indeed, there are examples of such differences in the literature. Whereas mouse β-cells depend on Glut2 for transmembrane glucose transport, human β-cells instead depend on Glut1 (36). These discrepancies might also be relevant to the understanding of the genetics of type 2 diabetes. Clearly, a gene polymorphism associated with increased risk of type 2 diabetes is more likely to affect insulin secretion if the gene is expressed in human β-cells and vice versa.

This work is supported by the Wellcome Trust, the European Union (Biosim [LSHB-CT-2004-005137] and Eurodia [SHM-CT-2006-518153]), and the Department of Health (NIHR Biomedical Research Centres funding scheme).

We thank Dr. S. Hughes, Dr. D. Gray, and Dr. S. Cross for isolation of human islets and D. Wiggins for assistance with hormone release measurements.