Glucose Regulation of β-Cell K_ATP Channels: Is a New Model Needed?

The K_ATP channel plays a key role in coupling the metabolism of glucose to the activation of L-type calcium channels required for stimulation of insulin secretion rate. Fuels, most importantly glucose, are metabolized through the combined pathways of glycolysis, the tricarboxylate (TCA) cycle, and oxidative phosphorylation, thereby increasing the cytosolic ratio of ATP to ADP to close K_ATP channels (1). Models of β-cell glucose sensing and stimulation-secretion coupling need to consider both phases of glucose-induced secretory response: 1) the initial response to a postprandial step increase in glucose (first phase; 2–5 min) and 2) sustained oscillations, characterized by bursts of plasma membrane depolarization (1- to 5-min duration) at high glucose (second phase).

In the complete oxidation of glucose, the ATP generated by pyruvate metabolism in the TCA cycle followed by oxidative phosphorylation is quantitatively the largest source of ATP and dominates overall production (2). Nevertheless, the close relationship between glucokinase, glycolysis, and insulin secretion suggests that reactions in the glycolytic pathway have a higher control strength for insulin secretion than would be expected based on the relatively small contribution to total ATP production (3). However, the enzymatic configuration of the β-cell (reviewed in Rutter et al. [1]), characterized by low levels of lactate dehydrogenase and essentially zero plasma membrane lactate/pyruvate transporter (MCT-1) activity, allows these cells to metabolize >85% of glucose carbon to CO₂ and H₂O. Thus, glycolysis and mitochondrial respiration are intimately linked and coordinately controlled: inhibition of the latter blocks the former, in contrast to most other cell types.

The total ATP generation rate in β-cells from glycolysis includes two molecules of ATP/glucose and five molecules of ATP/glucose from glycolytically derived NADH that are then shuttled into the mitochondria and oxidized in the electron transport chain via the glycerol phosphate shunt, which is particularly active in these cells (1). Three sources of ATP are thus involved (Fig. 1A): from glycolysis, from glycolytically derived NADH, and from the metabolism of pyruvate within mitochondria (Fig. 1A). Respectively, these yield 2, 5, and 30 molecules of ATP/glucose (4). For glycolytic ATP production to be an important determinant of K_ATP channel closure, this source would need to overcome the much higher production of ATP from pyruvate-driven oxidative mitochondrial metabolism. Quantitative modeling (4) also indicates that the mitochondrial contribution is relatively constant as glucose increases.

Four scenarios have been suggested that would enable glycolytic ATP to control K_ATP as glucose rises, in the face of the much higher rates of production of ATP by mitochondrial oxidation.

This scenario was supported by studies where hydrogen shuttle activity on glycolytically generated NADH was directly linked to glucose-stimulated K_ATP closure (5). Conversely, increases in glucose oxidation were counterbalanced by decreases in oxidation of endogenous fuels so that increased oxygen consumption rate (OCR) was wholly explained by glycolytic ATP and NADH oxidation (4). Thus, the scenario where the seven ATP molecules derived from glycolysis and mitochondrial shuttling of NADH makes up the predominant contribution to ATP/ADP levels is bioenergetically plausible. Whether this is the case when only 2 ATP molecules are derived from glycolysis, versus 35 from mitochondria, is unclear.

To explore these possibilities experimentally, Merrins and colleagues (6) have used genetic and pharmacological approaches to modulate pyruvate kinase (PK) activity. While PK activators enhanced glucose-stimulated insulin secretion (6), these agents also increased depolarization-stimulated exocytosis in capacitance recordings, hinting at a metabolism-independent effect. When floxed alleles of PKm1 or PKm2 were deleted selectively in the β-cell, loss of either PK isoenzyme was >90% as assessed in islet lysates (probably an underestimate given contamination from nontargeted islet cells). As expected, the effects of phosphoenolpyruvate (PEP) on K_ATP channel closure (see below) were abolished in the absence of a PK activator. However, glucose-stimulated Ca²⁺ dynamics and insulin secretion from isolated islets were largely unaffected in these models, and animals were metabolically normal (7). We note that these experiments may be complicated by the loss of histone kinase activity of PKm2 (8); broader changes at the transcriptomic or proteomic level were not assessed. In sum, these data do not support a critical and unique role for PK in controlling K_ATP channel closure.

Past studies have shown that there is rapid equilibrium between cytosolic and mitochondrial ATP and ADP pools (9). However, Merrins and colleagues (6) propose that microdomains (or nanodomains) of high ATP/ADP in the vicinity of the K_ATP channel are created by glycolytically generated ATP (from PK and phosphoglycerate kinase), alongside ATP provided by the action of intramitochondrial phosphoenolpyruvate carboxykinase (PEPCK; gene PCK2) and mitochondrial export of PEP (Fig. 1B). In support of this view, added PEP closed K_ATP channels in excised patches at a fixed ratio of ATP to ADP (6). This was attributed to a glycolytic metabolon that included PK (10). In this scenario, PEP exported from the mitochondria enters a microdomain, while ATP exported from the mitochondria does not, a condition that seems unlikely in the absence of an active barrier against such changes (see below). Moreover, the degree to which a privileged association between PK and K_ATP channels (10) exists in the living cell is unclear, since the patches examined may conceivably recede to form “Ω structures” that trap soluble cytosolic proteins but exclude mitochondria.

Comparisons of the effects of fuels metabolized exclusively by mitochondria with the effects of glucose have been used in the past to argue for the actions of mitochondrially derived ATP (11). Ketoisocaproate, glutamine, and leucine all provide substrate for the TCA cycle, increase OCR and calcium influx, and stimulate insulin secretion. Merrins and colleagues (6) propose that these fuels can support PEP generation by mitochondria, which would then generate ATP by the action of PK. Although PEPCK (PCK2) levels in the islet were previously thought to be low (12), Pck2 mRNA is detectable in purified mouse β-cells (https://huisinglab.com/data-ghrelin-ucsc/index.html), so that flux from oxaloacetate to PEP, although not measured, is plausible. As such, the above experiments involving fuels for the TCA cycle cannot definitively dissect the exclusive role of ATP generated from the action of PK on PEP from that of mitochondrially generated ATP on K_ATP channel closure. Nonetheless, unphosphorylated glyceraldehyde bypasses glycolysis and directly increases mitochondrial redox state (as reflected by cytochrome c reduction), OCR, intracellular calcium, and insulin secretion similarly to glucose, but it does so in the face of decreased NADH (reflecting the lack of involvement of the glycolytic enzyme glyceraldehyde phosphate dehydrogenase) (Fig. 2). Under these conditions, mitochondrially generated ATP must have free access to K_ATP channels without glycolytic stimulation (13). Thus, mitochondrial ATP generation is necessary and sufficient for glucose-stimulated insulin secretion.

Merrins and colleagues (6) suggest that as ATP/ADP increases from the metabolism of glucose (in the first phase), ADP falls to levels below those that permit oxidative phosphorylation. In tissues such as muscle, ATP synthesis through oxidative phosphorylation is regulated by increased demand (ADP supply), i.e., “pull.” Mitochondria within most cells are under dual control by both electron and ADP supply (14), and abundant evidence exists that this is the case in the β-cell (15). However, at basal (interprandial) glucose levels, the cytosolic ATP/ADP ratio in the β-cell is low (∼15 vs. ∼100 in most cells), and mitochondria are in “state 3.5” (15), i.e., OCR is not chiefly limited by ADP (though this exerts significant control) (6). Hence, there is no requirement to increase ADP supply, which in fact falls as cytosolic ATP/ADP increases. Consistent with this push control of OCR in β-cells, elevated glucose increases cytochrome c reduction, OCR, and calculated ATP/ADP ratio in parallel (16). Thus, ADP levels do not fall enough to substantially limit the operation of oxidative phosphorylation during the initial, large changes in metabolism in response to a step increase in glucose (first phase). In contrast, glibenclamide (provoking membrane depolarization and ATP consumption) increases OCR, without an increase in cytochrome c, and lowers ATP/ADP.

Further supporting the view above that K_ATP channels are regulated by mitochondrial bioenergetics, the time course of OCR changes during glucose challenge aligns closely with those of cytosolic ATP/ADP and K_ATP channel closure (Fig. 3A) (17). Thus, when single mouse islets are exposed to stepwise increases in glucose from 3 to 10 mmol/L, OCR reaches an initial peak, just before action potential firing and increased Ca²⁺ influx, whereupon a further increase in OCR occurs. In a later independent study (18), these changes were matched by an initial increase in NAD(P)H, which was further augmented as Ca²⁺ rose for the first time in the cytosol (Fig. 3B).

The slowing and then acceleration in OCR (Fig. 3A, asterisk) and NAD(P)H autofluorescence (Fig. 3B, asterisk) of the response to glucose likely reflects a combined effect of 1) increased ADP supply (mitochondria possibly approaching state 4 and then transitioning back to state 3.5 when cytosolic Ca²⁺ increases; see below) and 2) the stimulation by Ca²⁺ of intramitochondrial dehydrogenases (pyruvate, isocitrate, and 2-oxoglutarate dehydrogenase) (19). The importance of mitochondrial Ca²⁺ uptake is demonstrated by the effects of deleting the mitochondrial Ca²⁺ uniporter MCU (20). This pivotal role of Ca²⁺ is strong evidence for the involvement of mitochondria in controlling K_ATP channels.

In the second phase, elevated baseline changes in NAD(P)H (period of 3–6 min) were also evident and were entrained to the Ca²⁺ oscillations (Fig. 3C). The temporal correlation between mitochondrially derived NADH and OCR with K_ATP channel closure thus supports the involvement of mitochondrial substrate oxidation as the driver for insulin secretion while eclipsing the ATP generated by glycolysis.

When measured simultaneously in the same individual mouse β-cell, K_ATP channel closure (assessed via input resistance) in response to 17 mmol/L glucose aligned almost perfectly in time with the increase in bulk cytosolic ATP/ADP ratio (measured with the recombinant probe Perceval) (20). K_ATP channel closure was detectable when cytosolic ATP/ADP ratios had risen by <10% (Fig. 4A). During the subsequent electrical bursting and Ca²⁺ increases, membrane potential oscillations and input resistance nearly perfectly mirrored the bulk cytosolic ATP/ADP ratio, consistent with regulation of the former by the latter (Fig. 4A). Importantly, the dynamics of plasma membrane–proximal (within ∼70 nm) ATP/ADP changes, measured using near-field microscopy, are essentially identical to the bulk changes described above (21), arguing against a discrete ATP/ADP microdomain in this region.

Notably, while ATP/ADP ratios fluctuate in these experiments, they remain well above baseline levels observed at 3 mmol/L glucose (20,21) (Fig. 4A). The notion that ADP is too low (and ATP/ADP too high) to allow ATP generation by mitochondria during the silent phase between electrical bursts (dubbed Mito_Cat) (6) is challenged by the measurements described above (Fig. 4A) and similar recordings (22) at 10 mmol/L glucose (Fig. 4B and C). In both cases, cytosolic ATP/ADP increases gradually prior to the next electrical burst (Fig. 4B) and Ca²⁺ pulse (Fig. 4C). As Ca²⁺ falls between bursts, the parallel increase in ATP/ADP presumably reflects lowered ATP consumption for ion pumping (by Ca²⁺ pumps, Na⁺/K⁺ ATPase, etc.) (23) in the face of constant or lowered ATP synthesis. A new electrical burst occurs at the attainment of sufficient ATP/ADP for K_ATP channel closure. Although Mito_Ox and Mito_Cat/Synth (6) phases (i.e., oxidative and catabolic/synthetic phases) may thus exist, we prefer to define a maximum oxidative phosphorylation (Ox Phos^max) state (Fig. 4D and E) that bisects these. Importantly, there is no requirement in this model, or direct evidence for, the local generation of ATP from PEP in the Mito_Cat phase. In fact, this would blunt K_ATP channel reopening and membrane repolarization, suppressing oscillations.

Theoretical considerations and experimental evidence (10), as discussed here, provide only limited support for a new model of K_ATP channel regulation via glycolytic intermediates and/or mitochondrial cataplerosis but rather support the existing, simpler model based on mitochondrial respiratory chain activity. In striking contrast to the canonical model (1), little genetic evidence exists currently to corroborate the new model.

In our view, the following are needed to provide a more compelling case for the new model:

• 1.

  Demonstration in intact β-cells (e.g., by electron microscopy, energy transfer, or proximity assays) of the existence of a stable K_ATP channel–associated metabolon capable of excluding mitochondrially generated ATP;

• 2.

  A molecular definition of the protein-protein interactions that recruit PK isoforms to a K_ATP-proximal metabolon;

• 3.

  Disruption, possibly based on the work of Stryer (2), of these interactions, of the hypothesized K_ATP-proximal ATP/ADP microdomains, and of normal insulin secretion; and

• 4.

  Genetic evidence in humans or other animals.

During final review of this article, it was reported that the application of PEP plus ADP failed to close β-cell K_ATP channels (24), further emphasizing the need for experimental validation of the PK hypothesis.

Funding. G.A.R. was supported by a Wellcome Trust Investigator award (212625/Z/18/Z), UKRI-Medical Research Council program grant (MR/R022259/1), a National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, project grant (R01DK135268), a Canadian Institutes of Health Research–JDRF Team grant (CIHR-IRSC TDP-186358 and JDRF 4-SRA-2023-1182-S-N), Centre de Recherche du CHUM start-up funds, and an Innovation Canada John R. Evans leader award (CFI 42649). I.R.S. was supported by grants from the National Institutes of Health (R01 GM148741) and the Leona M. and Harry B. Helmsley Foundation (no. FA218303).

Duality of Interest. G.A.R. has received funding from and is a consultant for Sun Pharmaceuticals, Inc. I.R.S. has financial ties to EnTox Sciences, Inc. (Mercer Island, WA). No other potential conflicts of interest relevant to this article were reported.