β-cell deletion of the PKm1 and PKm2 isoforms of pyruvate kinase in mice reveals their essential role as nutrient sensors for the KATP channel

Pyruvate kinase (PK) and the phosphoenolpyruvate (PEP) cycle play key roles in nutrient-stimulated KATP channel closure and insulin secretion. To identify the PK isoforms involved, we generated mice lacking β-cell PKm1, PKm2, and mitochondrial PEP carboxykinase (PCK2) that generates mitochondrial PEP. Glucose metabolism was found to generate both glycolytic and mitochondrially derived PEP, which triggers KATP closure through local PKm1 and PKm2 signaling at the plasma membrane. Amino acids, which generate mitochondrial PEP without producing glycolytic fructose 1,6-bisphosphate to allosterically activate PKm2, signal through PKm1 to raise ATP/ADP, close KATP channels, and stimulate insulin secretion. Raising cytosolic ATP/ADP with amino acids is insufficient to close KATP channels in the absence of PK activity or PCK2, indicating that KATP channels are primarily regulated by PEP that provides ATP via plasma membrane-associated PK, rather than mitochondrially derived ATP. Following membrane depolarization, the PEP cycle is involved in an ‘off-switch’ that facilitates KATP channel reopening and Ca2+ extrusion, as shown by PK activation experiments and β-cell PCK2 deletion, which prolongs Ca2+ oscillations and increases insulin secretion. In conclusion, the differential response of PKm1 and PKm2 to the glycolytic and mitochondrial sources of PEP influences the β-cell nutrient response, and controls the oscillatory cycle regulating insulin secretion.


INTRODUCTION 65
Maintenance of euglycemia relies on β-cells to couple nutrient sensing with appropriate 66 insulin secretion. Insulin release is stimulated by the metabolism-dependent closure of ATP-67 sensitive K + (K ATP ) channels (Ashcroft et al., 1984;Cook and Hales, 1984;Misler et al., 1986; 68 Rorsman and Trube, 1985), which triggers Ca 2+ influx and exocytosis (Anderson and Long,69 1947; Grodsky et al., 1963). Contrary to what is often believed, the glucose-induced signaling 70 process in β-cells has not been largely solved, and the entrenched model implicating a rise in 71 mitochondrially-derived ATP driving K ATP channel closure (Campbell and Newgard, 2021;72 Prentki et al., 2013) is incomplete and possibly wrong, the main reason being that it does not 73 consider other sources of local ATP production that may be key for signaling (Corkey, 2020;74 Lewandowski et al., 2020; Merrins et al., 2022). The recent discovery that pyruvate kinase (PK), 75 which converts ADP and phosphoenolpyruvate (PEP) to ATP and pyruvate, is present on the β-76 cell plasma membrane where it is sufficient to raise sub-plasma membrane ATP/ADP 77 (ATP/ADP pm ) and close K ATP channels (Lewandowski et al., 2020) provides an alternative 78 mechanism to oxidative phosphorylation for K ATP channel regulation. Based on this finding, 79 Lewandowski PK progressively increases the ATP/ADP ratio, and by lowering ADP slows oxidative 90 phosphorylation. The shift to a higher mitochondrial membrane potential (ΔΨ m ) elevates the 91 NADH/NAD + ratio in the mitochondrial matrix and slows the TCA cycle, increasing acetyl-CoA 92 that allosterically activates pyruvate carboxylase, the anaplerotic consumer of pyruvate that fuels 93 oxaloacetate-dependent PEP synthesis by mitochondrial PEP carboxykinase (PCK2). The return 94 of mitochondrial PEP to the cytosol completes the "PEP cycle" that helps fuel PK, which raises 95 ATP/ADP pm to close K ATP channels. Following membrane depolarization and Ca 2+ influx, the 96 increased workload (ATP hydrolysis) associated with ion pumping and exocytosis elevates 97 cytosolic ADP, which activates oxidative phosphorylation to produce ATP that sustains insulin 98 secretion in a phase referred to as Mito Ox . An unresolved aspect of this model is whether plasma 99 membrane-compartmentalized PK activity is required to close K ATP channels. This question is 100 important because in the current canonical model of fuel-induced insulin secretion, an increase in 101 the bulk cytosolic ATP/ADP ratio (ATP/ADP c ) is generally assumed to close K ATP channels. 102 In the Mito Cat -Mito Ox model, PK has two possible sources of PEP that may differentially 103 regulate K ATP closure: glycolytic PEP produced by enolase, and mitochondrial PEP produced by 104 PCK2 in response to anaplerosis (Figure 1a). About 40% of glucose-derived PEP is generated by 105 PCK2 in the PEP cycle and is closely linked to insulin secretion (Abulizi et al., 2020; Jesinkey et 106 al., 2019; Stark et al., 2009). However, it remains unclear how the PEP cycle influences glucose-107 stimulated oscillations. We hypothesize that mitochondrial PEP derived from PCK2 may provide 108 a glycolysis-independent mechanism by which PK rapidly increases ATP/ADP pm locally at the 109 K ATP channel in response to amino acids, which are potent anaplerotic fuels. 110 The isoforms of PK, each with different activities and mechanisms of control, may 111 differentially regulate K ATP channels (Figure 1a). β-cells express the constitutively-active PKm1 112 as well as two allosterically-recruitable isoforms, PKm2 and PKL, which are activated by 113 glycolytic fructose-1,6-bisphosphate (FBP) generated upstream by the phosphofructokinase which due to its constitutive (FBP-insensitive) activity might be ideal in situations of high 120 oxidative workload, as in cardiac myocytes (Li et al., 2021). β-cells may shift their reliance upon 121 different PK isoforms throughout the oscillatory cycle, as the levels of glycolytic FBP rise during 122 Mito Cat and fall during Mito Ox (Lewandowski et al., 2020;Merrins et al., 2016Merrins et al., , 2013. 123 Here, we show that PK is essential for K ATP closure -amino acids that effectively raise 124 ATP/ADP c cannot close K ATP channels without PK. We further demonstrate that both PKm1 and 125 PKm2 are active in the K ATP channel microcompartment with at least two required functions. 126 First, spatial privilege provides redundancy in the β-cell glucose response, by permitting the 127 minor PKm2 isoform, when activated by FBP, to transmit the signal from glucose to K ATP 128 despite contributing only a small fraction of the whole cell PK activity. Second, the composition 129 of PK isoforms within the K ATP compartment tunes the β-cell response to amino acids, which 130 provide mitochondrial PEP for PKm1 without also generating the FBP needed to allosterically 131 activate PKm2. Using β-cell PCK2 deletion, we found that mitochondrially-derived PEP signals 132 to the plasma membrane PK-K ATP microcompartment during Mito Cat  increased by 29% (Figure 1c). This partial compensation is expected since PKm1 and PKm2 are 146 alternative splice variants of the Pkm gene (Li et al., 2021;Israelsen et al., 2013). We generated 147 Pck2 f/f mice (Figure 1-figure supplement 2 Figure 1d). None of these knockout mice were glucose intolerant (Figure 1e-g), nor did 151 they exhibit changes in meal tolerance by oral gavage (Figure 1-figure supplement 3a-c). 152 The contributions of each PK isoform relative to total PK activity was determined in the 153 islet lysates. decreased by 97% compared to controls (Figure 1h), and was too low to estimate K m accurately. 161 The residual PK remained sensitive to activation by both FBP and PKa, identifying an 162 allosterically-recruitable PK pool that accounts for only about 10% of the PK activity present in 163 control islets (Figure 1h). Conversely, β-cell PKm2 deletion lowered islet lysate PK V max by 164 ~20% in the absence of activators, and eliminated both the K m and V max response to PKa and 165 FBP (Figure 1i), thus ruling out any measurable PKL activity. Taken together, mouse islet PK 166 activity is composed of >90% PKm1, with a variable contribution from PKm2 depending on the 167 FBP level. If only considered in terms of total cellular activity related to nutrient-induced insulin 168 secretion (i.e. in the absence of any compartmentalized functions), PKm1 should be dominant 169 over PKm2 under all physiologic conditions. The fact that PKm1-βKO mice maintain metabolic 170 health with unaltered glucose tolerance into adulthood suggests that the remaining PK activity is 171 sufficient for β-cell function, and led us to hypothesize that both PKm1 and PKm2 function in 172 the K ATP channel microcompartment.

173
Both PKm1 and PKm2 are associated with the plasma membrane and locally direct K ATP 174 channel closure, however PKm2 requires allosteric activation even at high PEP levels 175 We previously demonstrated that PEP, in the presence of saturating ADP concentrations, 176 can close K ATP channels in mouse and human β-cells (Lewandowski et al., 2020). This suggests 177 that PK is present near K ATP and locally lowers ADP and raises ATP to close K ATP channels. 178 Excised patch-clamp experiments, which expose the inside of the plasma membrane to the bath 179 solution (i.e. the inside-out mode), provide both the location of endogenous PK as well as its 180 functional coupling to K ATP channels in native β-cell membranes. This approach was applied in 181 combination with β-cell deletion of PKm1 and PKm2 to directly identify the isoforms of the 182 enzyme present in the K ATP microdomain. K ATP channels were identified by inhibition with 1 183 mM ATP, which blocked the spontaneous opening that occurs after patch excision (Figure 2a). 184 Channel activity was restored using a test solution containing 0.5 mM ADP and 0.1 mM ATP. In 185 control β-cells, the further addition of 5 mM PEP closed K ATP , as shown by a 77% reduction in 186 the total power (a term reflecting both the frequency and channel open time) (Figure 2a), 187 compared with only a 29% reduction in PKm1-βKO cells (Figure 2b). Note that K ATP channel 188 closure occurred in control β-cells despite the continuous deluge of the channel-opener MgADP. 189 Thus, it is not the PEP itself, but the PK activity in the K ATP microcompartment that is 190 responsible for K ATP closure.

191
To test for a role of PKm2 in the K ATP microcompartment, PKm1-βKO cells were 192 preincubated in the presence of 10 μM PKa, which restored PEP-dependent K ATP channel closure 193 to the same extent as the control (1 mM ATP) ( Figure 2c). PKa had a similar effect when applied 194 acutely (Figure 2d), indicating that PKm2 does not require allosteric activation to localize to the 195 plasma membrane. Although the PEP concentration is estimated to be 1 mM in rat islets (Sugden 196 and Ashcroft, 1977), 5 mM PEP was chosen to exceed the K m of PKm2 in the absence of FBP 197 (Lewandowski et al., 2020). Therefore, the response of PKm2 to PKa at high PEP levels 198 indicates that K ATP closure requires an additional increase in the V max via either allosteric 199 activation of individual subunits of PKm2, or perhaps more likely, that the functional interaction 200 is sensitive to the quaternary structure of PKm2. β-cells lacking PKm2 maintained channel 201 closure with the same power as 1 mM ATP, which is attributable to the sufficiency of 202 endogenous PKm1 (Figure 2e). Thus, metabolic compartmentation of PKm1 and PKm2 to the 203 plasma membrane provides a redundant mechanism of K ATP channel regulation when PKm2 is 204 allosterically activated, as well as a compelling explanation for the ability of PKm1-βKO mice to 205 tolerate a near-complete loss of β-cell PK activity (Figure 1e,h).

206
PKm1 and PKm2 are redundant for glucose-dependent Ca 2+ influx 207 The rescue of PKm1 deficiency by PKa in the K ATP microcompartment (Figure 2c-d) 208 suggests that PKm1 and PKm2 exert shared control over K ATP closure, provided that glucose is 209 present to generate FBP to activate PKm2. To test this further we examined Ca 2+ dynamics with 210 FuraRed while using a near-infrared dye, DiR, to facilitate simultaneous imaging of PKm1-, 211 PKm2-, and PCK2-βKO islets with their littermate controls ( Figure 3a). β-cell deletion of PKm1 212 revealed no difference in the oscillatory period or amplitude and a modest increase in the fraction 213 of each oscillation spent in the electrically-active state (i.e. the duty cycle) (Figure 3b). β-cell 214 PKm2 deletion increased the period of Ca 2+ oscillations, while having no impact on the 215 amplitude or the duty cycle ( Figure 3d). In addition, PKm1 and PKm2 knockouts had no 216 discernable difference in first-phase Ca 2+ parameters (i.e., time to depolarization, amplitude, and 217 duration of first phase) following an acute rise in glucose from 2 to 10 mM (Figure 3c,e). These 218 data confirm in situ that PKm1 and PKm2 are largely redundant at high glucose.

219
PK and the PEP cycle are implicated in both on-and off-switches for Ca 2+ influx 220 To study the interaction of PK with the PEP cycle, we performed islet Ca 2+ 221 measurements using PK activators and PCK2-βKO islets, in the latter case using islet barcoding 222 to simultaneously image islets isolated from littermate controls. Consistent with the ability of 223 allosteric PKm2 activation to accelerate K ATP closure, acute application of PKa to wild-type 224 islets reduced the period as well as the amplitude of the steady-state glucose-induced Ca 2+ 225 oscillations ( 1b), leading to a modest reduction in the duty cycle as well as a more significant reduction in the 228 period of the oscillation (Figure 3f). These observations suggest that the PKm2-driven PEP cycle 229 regulates the onset, and even more strongly, the termination of Ca 2+ influx. Consistently, in 230 PCK2-βKO islets where mitochondrial PEP production is inhibited, both the period and 231 amplitude of glucose-stimulated Ca 2+ oscillations were increased relative to controls islets 232 (Figure 3h). Although the duty cycle also increased, it was only by a small margin. The period 233 lengthening occurred in part from an increased duration of Mito Cat , and especially from an 234 increased duration of Mito Ox (Figure 3-figure supplement 1c). Taken together, these data indicate 235 that PKm2 controls both an "on-switch" and an "off-switch" for Ca 2+ oscillations, both of which 236 depend on the mitochondrial production of PEP.

237
While above experiments examined conditions at a fixed elevated glucose concentration 238 (10 mM), we also investigated Ca 2+ dynamics following the transition from low to high glucose 239 where first-phase insulin secretion is observed. Preincubation of control islets with PKa reduced 240 the time to depolarization as well as the duration of the first-phase Ca 2+ influx ( Figure 3g). 241 Conversely, depolarization was delayed in PCK2-βKO islets ( Figure 3i). In this case, the 242 duration of the first-phase Ca 2+ pulse was not calculated since nearly 60% of PCK2-βKO islets 243 failed to exit the first phase plateau in order to begin oscillations, as compared with only 27% of 244 control islets (Figure 3i). In other words, while the PCK2 knockout had a weaker first phase Ca 2+ 245 rise, it had a much longer plateau that failed to turn off effectively. Hence, PKm2 activation and 246 PCK2 serve as on-switches for promoting glucose-stimulated Ca 2+ influx during the triggering 247 phase (Mito Cat ), and with a quantitatively larger effect, off-switches during the secretory phase 248 (Mito Ox ).

249
Mitochondrial PEP carboxykinase (PCK2) is essential for amino acids to promote a rise in 250 cytosolic ATP/ADP 251 To determine whether PCK2, PKm1, or PKm2 are essential for the rise in the cytosolic 252 ATP/ADP ratio generated by high glucose or amino acids, we used β-cell specific expression of 253 Perceval-HR biosensors to measure ATP/ADP c . We found that, as with Ca 2+ , there were no 254 significant differences in glucose-stimulated ATP/ADP c detected in the PKm1-or PKm2-βKO 255 islets (Figure 4a,b), demonstrating the redundancy of the two isoforms at high glucose for 256 ATP/ADP c generation. Similarly, we detected no significant difference in glucose-stimulated 257 ATP/ADP c in islets from the PCK2-βKO ( Figure 4c). 258 Amino acids (AA) are obligate mitochondrial fuels that simultaneously feed oxidative 259 and anaplerotic pathways. AA can be used as a tool for separating mechanistic components of 260 the secretion mechanism because at low glucose they can, independently of glycolysis, raise 261 ATP/ADP c and elicit K ATP channel closure, Ca 2+ influx, and insulin release. In particular, 262 glutamine and leucine generate PEP via glutamate dehydrogenase (GDH)-mediated anaplerosis 263 that is followed by PCK2-mediated cataplerosis of PEP (Kibbey et al., 2014;Stark et al., 2009). 264 We first examined whether restriction of mitochondrial PEP production in PCK2-βKO islets 265 impacts the cytosolic ATP/ADP c ratio. To limit glycolytic PEP, islets were incubated at 2.7 mM 266 glucose. The islets were then stimulated with a mixture of AA including leucine and glutamine to 267 allosterically activate and fuel GDH, respectively. Consistent with defective PEP cataplerosis, 268 the ATP/ADP c response of PCK2-βKO islets was only 44% of control islets in response to AA 269 ( Figure 4d). In this setting of PCK2 depletion, pharmacologic PK activation did not recover any 270 of the AA-induced ATP/ADP c response due to the absence of either a glycolytic or 271 mitochondrial PEP source ( Figure 4g). Comparatively, deletion of either PKm1 or PKm2 had 272 only modest effects on the β-cell ATP/ADP c response to AA (Figure 4e,f). However, PKa 273 completely recovered the AA-induced rise in ATP/ADP c in PKm1-βKO islets, in which the 274 allosteric PKm2 isoform remains ( Figure 4h). As expected, PK activation had no effect in 275 PKm2-βKO islets (Figure 4i), confirming an on-target effect of TEPP-46.

276
Mitochondrial fuels that stimulate a bulk rise in ATP/ADP c fail to close K ATP in the absence of 277 PK 278 Mitochondria are located throughout the β-cell, including near the plasma membrane, 279 where submembrane ATP microdomains have been observed (Griesche et al., 2019; Kennedy et 280 al., 1999). Since PK is localized to the plasma membrane, we wondered whether during AA 281 stimulation mitochondria can provide PEP to facilitate PK-dependent K ATP closure, or 282 alternatively, whether mitochondria can serve as a direct source of ATP for K ATP channel 283 closure. To determine whether mitochondrial PEP impacts the K ATP channel microcompartment, 284 we monitored K ATP channel currents in intact β-cells in the cell-attached configuration in 285 response to bath-applied AA at 2.7 mM glucose ( Figure 5a). Mixed AA with or without PKa 286 reduced K ATP channel power (reflecting the total number of transported K + ions) by ~75% in 287 control β-cells ( Figure 5b). However, no K ATP closure was observed in the absence of PCK2, 288 even with PK activator present ( Figure 5c). These findings indicate that mitochondrially-derived 289 PEP can signal to the K ATP channel microcompartment, and is essential for K ATP closure in 290 response to AA. 291 Unlike PCK2-deficient β-cells, PKm1-deficient β-cells stimulated with AA were capable 292 of increasing ATP/ADP c to a similar level as control and PKm2-deficient β-cells ( Figure 4). This 293 model provides a unique opportunity to directly test the canonical model of fuel induced insulin 294 secretion, where a rise in ATP/ADP in the bulk cytosol is thought to be sufficient to close K ATP 295 channels. While K ATP channels were efficiently closed by AA in control and PKm2-deficient β-296 cells (Figure 5b and 5d), K ATP channels failed to close in β-cells lacking PKm1 ( Figure 5e). As in 297 excised patches (Figure 2c-d), pharmacologic activation of PKm2 was sufficient to rescue K ATP 298 closure in PKm1-deficient β-cells ( Figure 5f). These findings demonstrate that PK activity is 299 essential for K ATP channel closure in response to AA, and argue strongly against the canonical 300 model in which mitochondrially-derived ATP raises ATP/ADP c to close K ATP channels. 301 Mitochondrially-derived PEP drives K ATP closure and Ca 2+ influx during Mito Cat , and 302 accelerates K ATP reopening and Ca 2+ extrusion during Mito Ox 303 As for glucose, AA-stimulated Ca 2+ influx follows a distinct triggering and secretory 304 phase, representing Mito Cat and Mito Ox , respectively ( Figure 5g). Following PK-dependent K ATP 305 channel closure with AA, the ability of PKa to partially reopen K ATP channels ( Figure 5b) 306 suggests that PKm2 might also serve as an "off-switch" during Mito Ox that eventually increases 307 the channel opening by activating the PEP cycle. This concept is consistent with the with the 308 ability of PKa to both hasten the onset and shorten the duration of Ca 2+ pulses (Figure 3f,g). It is 309 also consistent with the observation that mitochondrially-derived PEP is necessary to switch off 310 glucose-dependent Ca 2+ influx, as shown in PCK2-βKO islet experiments (Figure 3h,i). In other 311 words, the PEP cycle would have a dual function in the β-cell: during Mito Cat , the PEP cycle 312 facilitates K ATP closure and Ca 2+ influx; during Mito Ox , the PEP cycle may facilitate K ATP 313 channel reopening, Ca 2+ extrusion and turn off insulin secretion. To test this concept further, we 314 examined the effect of PKa on AA-stimulated Ca 2+ influx. When applied before AA stimulation, 315 during Mito Cat , PKa increased Ca 2+ influx ( Figure 5h). By contrast, PKa application after the 316 initial Ca 2+ rise, during Mito Ox , reduced cytosolic Ca 2+ ( Figure 5i). These data confirm two 317 temporally separated functions of PK and mitochondrial PEP -representing on-and off-switches 318 for β-cell Ca 2+ . While the on-switch is mediated by PK-dependent closure of K ATP channels, the 319 cellular location of the mitochondrial PEP-dependent off-switch is presently unclear. 320 PKm1 and PKm2 respond differentially to the glycolytic and mitochondrial sources of PEP 321 We next examined the functional consequence of β-cell PKm1, PKm2, and PCK2 322 deletion on AA-stimulated Ca 2+ influx and insulin secretion at both low and high glucose ( Figure  323 6). In PCK2-βKO islets at 2 mM glucose, the AA-induced Ca 2+ response was reduced along with 324 insulin secretion (Figure 6a,b). In the presence of 10 mM glucose, β-cell PCK2 deletion did not 325 impact insulin secretion because of the restored glycolytic PEP supply (Figure 6c). In the 326 presence of high glucose and leucine to maximally stimulate anaplerosis, PCK2-βKO islets fail 327 to inactivate during Mito Ox, as indicated by the sustained Ca 2+ plateau (Figure 3h,i). Under 328 similar conditions, when insulin secretion was stimulated by 10 mM glucose and mixed AA, 329 insulin secretion was higher in PCK2-βKO islet than controls ( Figure 6c). 330 Like the PCK2-βKO, the AA-stimulated Ca 2+ and secretory responses of PKm1-βKO 331 islets were blunted at 2 mM glucose (Figure 6d,e). However, the insulin secretory response 332 remained intact at 10 mM glucose in PKm1-βKO islets, whether or not AA were present ( Figure  333 6f), due to the sufficiency of PKm2 in the presence of glycolytic FBP. These findings are entirely 334 consistent with the differential response of PKm1-βKO islets to AA vs. glucose stimulation 335 observed in the Ca 2+ and K ATP channel recordings above (Figures 3 and 5).

336
Similarly to PCK2, allosterically-activated PKm2 has dual functions during Mito Cat and 337 Mito Ox , as evidenced by the Ca 2+ and K ATP channel measurements shown in Figures 3 and 5. 338 Both the AA-induced Ca 2+ response and insulin secretion were greatly increased in PKm2-βKO 339 islets compared to controls (Figure 6g,h). As expected, while PK activation increased AA-340 stimulated Ca 2+ release in control islets ( Figure 6-figure supplement 1a), it had no effect in 341 PKm2-βKO islets ( Figure 6-figure supplement 1b), confirming an on-target effect of TEPP-46. 342 Glucose alone, but especially in combination with AA, stimulated enhanced secretion in the 343 absence of PKm2 (Figure 6i). Thus, while either PKm1 or PKm2 is sufficient to initiate glucose-344 stimulated insulin secretion during Mito Cat , PKm2 is essential for mitochondrial PEP to switch 345 the system off during Mito Ox . When PKm2 was deleted, and PKm1 increased (Figure 1c), the 346 system is shifted towards greater nutrient-induced and PCK2-dependent insulin secretion with 347 heightened sensitivity to anaplerotic fuels. 348 These data provide genetic evidence that PEP has metabolic control over PK-dependent 351 ATP/ADP pm generation, K ATP closure, Ca 2+ signaling, and insulin secretion. β-cell PK isoform 352 deletion experiments demonstrate that plasma membrane-associated PK is strictly required for 353 K ATP channel closure and provide rigorous genetic evidence for the Mito Cat -Mito Ox model of 354 oscillatory metabolism and insulin secretion (Merrins et al., 2022). Our results indicate that PK is 355 controlled by two different sources of PEP -glycolytic and mitochondrial ( Figure 1a). Glucose 356 signals to PK via both glycolytic and mitochondrially-derived PEP, whereas amino acids signal 357 to PK exclusively through mitochondrially-derived PEP. Since amino acids do not generate FBP, 358 which is needed to allosterically activate PKm2, PKm1 is necessary to raise ATP/ADP pm , close 359 K ATP channels, and stimulate insulin secretion in response to amino acids. Furthermore, our work 360 supports the concept that the canonical model of fuel-induced insulin secretion, whereby 361 mitochondrially-derived ATP produced via the electron transport chain raises ATP/ADP c to 362 close K ATP channels, is possibly wrong. The evidence for this new view of β-cell metabolic 363 signaling is that while amino acids efficiently raise ATP/ADP c , they do not close K ATP channels 364 in the absence of plasma membrane PK activity. Finally, the data indicate that PK and the PEP 365 cycle have a dual role in the control of insulin secretion -they act as on-signals during the 366 triggering phase, Mito Cat , and as off-signals during the active secretory phase, Mito Ox . We 367 discuss each of these findings in the sections below.

368
A fuel-stimulated rise in ATP/ADP c was shown to be insufficient to close K ATP channels, 369 ruling out a key aspect of the "canonical model" in which accelerated mitochondrial metabolism 370 raises ATP/ADP c to close K ATP channels (Campbell and  increased ATP/ADP c similarly to control β-cells, but were unable to close K ATP channels. 373 Pharmacological activation of PKm2, present on the plasma membrane of PKm1-deficient β-374 cells, acutely restored K ATP closure. Thus, while mitochondrially-derived ATP/ADP c may help to 375 buffer the PK-dependent rise in ATPADP pm , PK is essential for K ATP closure in intact β-cells. 376 Plasma membrane-associated PK is also sufficient for K ATP  confirm the LoxP insertions around exon 5 (ENSMUSE00000399990; Figure 1- figure  481 supplement 2), and the intervening region was sequence confirmed with Sanger sequencing. 482 Pck2 islet preparations were passed through a sieve. Following washes, islet preparations were then 498 resuspended in 10 mL HBSS, and 10 mL of Lymphocyte Separation Medium 1077 (Sigma 499 Aldrich C-44010) was slowly added below the layer of islet preparation/HBSS using a 10 mL 500 syringe with an 18 G needle (Air-Tite ML1018112). Islet preparations were centrifuged at 800 501 rcf for 15 minutes then supernatant was poured into a new 50 ml conical and HBSS/BSA was 502 added to 50 ml. Islets were centrifuged for 5 minutes, supernatant was removed, and islets were 503 resuspended in 10-20 mL, poured into a petri dish then picked into dishes containing RPMI 1640 504 (Gibco #11875119).

506
Glucose and meal tolerance tests 507 Mice aged 11-22 weeks were fasted overnight for 16 hours prior to i.p. injection of glucose (1 508 g/kg body weight, prepared in PBS) or oral gavage of liquid Ensure (10 mL/kg). Blood glucose 509 was measured from the tail using a glucometer (Contour).

511
Cloning and adenoviral delivery of biosensors 512 Generation of adenovirus carrying genetically-encoded ATP/ADP biosensors (Perceval-HR) 513 under control of the insulin promoter was described previously . High-titer 514 adenovirus was added to islets immediately after islet isolation and incubated for 2 h at 37°C 515 then moved to fresh media. Imaging was performed 3 days post isolation.

517
Western blots 518 Islets were lysed using 0.1% Triton-X100 in phosphate buffered saline (1 µl/islet). Islets in lysis 519 buffer were incubated at room temperature for 15 minutes, vortexed for 30 seconds, 520 frozen/thawed, and vortexed for 30 seconds again. Cell lysate was spun down at max speed in a 521 Instruments) and polished by microforge (Narishige MF-830) to a final tip resistance of 5-10 574 MΩ. On-cell recording started after formation of a stable gigaseal (>2.5 GΩ) and inside-575 out recording started after withdrawal of the pipette and establishment of the excised inside-576 out configuration. A HEKA Instruments EPC10 patch-clamp amplifier was used for registration 577 of current. Data was Bessel filtered online at 1 kHz and single channel currents were analyzed 578 offline using ClampFit analysis module of pCLAMP 10 software (Molecular Devices).

597
Quantification and statistical analysis 598 SEM. Statistical significance was determined by two-way or one-way ANOVA with Sidak 600 multiple-comparisons test post hoc or Student's t-test as indicated in the figure legends. Data 601 were continuous and normally distributed so were analyzed with parametric tests. Differences 602 were considered to be significant at P < 0.05. Calculations were performed using GraphPad 603 Prism.

605
Materials availability statement 606 Pck2-floxed animals will be shared on a collaborative basis pending their availability. Inquiries 607 should be directed to the corresponding author.

757
(A) Hypothesized model in which PK in the K ATP microcompartment is fueled by two sources of 758 PEP-glycolytic PEP generated by enolase, and mitochondrial PEP generated by PCK2 in 759 response to anaplerotic fuels. β-cells express three isoforms of PK, constitutively active PKm1 760 and allosterically recruitable PKm2 and PKL that are activated by endogenous fructose 1,6-761 bisphosphate (FBP) or pharmacologic PK activators (PKa).