Prolyl-4-hydroxylase 3 maintains β cell glucose metabolism during fatty acid excess in mice

The α-ketoglutarate–dependent dioxygenase, prolyl-4-hydroxylase 3 (PHD3), is an HIF target that uses molecular oxygen to hydroxylate peptidyl prolyl residues. Although PHD3 has been reported to influence cancer cell metabolism and liver insulin sensitivity, relatively little is known about the effects of this highly conserved enzyme in insulin-secreting β cells in vivo. Here, we show that the deletion of PHD3 specifically in β cells (βPHD3KO) was associated with impaired glucose homeostasis in mice fed a high-fat diet. In the early stages of dietary fat excess, βPHD3KO islets energetically rewired, leading to defects in the management of pyruvate fate and a shift from glycolysis to increased fatty acid oxidation (FAO). However, under more prolonged metabolic stress, this switch to preferential FAO in βPHD3KO islets was associated with impaired glucose-stimulated ATP/ADP rises, Ca2+ fluxes, and insulin secretion. Thus, PHD3 might be a pivotal component of the β cell glucose metabolism machinery in mice by suppressing the use of fatty acids as a primary fuel source during the early phases of metabolic stress.

βPHD3KO islets is associated with impaired glucose-stimulated ATP/ADP rises, Ca 2+  The prolyl-hydroxylase domain proteins (PHD1-3) encoded for by the Egl nine homolog genes 56 are alpha ketoglutarate-dependent dioxygenases, which regulate cell function by catalyzing 57 hydroxylation of peptidyl prolyl residues within various substrates using molecular oxygen (1-58 4). There are three well-described mammalian isozymes: PHD1, PHD2 and PHD3, which were 59 originally described as hydroxylating the alpha subunit of the transcription factor Hypoxia-60 Inducible Factor (HIF) under normoxia (4), thus targeting it for polyubiquitylation and 61 proteasomal degradation. When oxygen concentration becomes limited, PHD activity 62 decreases and HIF is stabilized, leading to dimerization with the beta subunit and 63 transcriptional regulation of target genes regulating the cellular response to hypoxia (5). While 64 PHDs are generally regarded to be master HIF regulators, it is becoming increasingly apparent 65 that they target a range of other substrates influencing cell function (6-9). 66 PHD3 is unusual amongst the PHDs: it is transcriptionally regulated by HIF1 during 67 hypoxia (10), although it does not always act to destabilize HIF1 (11, 12). A number of roles 68 for PHD3 have been described under conditions of stress or hypoxia, including: macrophage 69 influx and neutrophil survival (13, 14), apoptosis in various cancer models (8,15,16), and 70 tumor cell survival (9) (reviewed in (17)). Due to the dependence of PHD3 on alpha-71 ketoglutarate and oxygen for its activity (18), many of these actions are likely to be mediated 72 through alterations in cell metabolism (19). Indeed, PHD3 increases glucose uptake in cancer 73 cells through interactions with pyruvate kinase M2 (8, 20). In tumors exhibiting mutations in 74 succinate dehydrogenase, fumarate hydratase and isocitrate dehydrogenase 1 and 2 (21-23), 75 PHD3 activity is altered by aberrantly high cytosolic concentrations of succinate, fumarate and 76 2-hydroxyglutarate (2-HG), suggesting that inactivation of this enzyme might be involved in 77 the cellular transformation process. PHD3 has more recently been shown to hydroxylate and 78 activate acetyl-CoA carboxylase 2 (ACC2), defined as the fatty acid oxidation gatekeeper, 79 thus decreasing fatty acid breakdown and restraining myeloid cell proliferation during nutrient 80 incretin release from the intestine (41), was similar in βPHD3CON and βPHD3KO mice ( Figure  137 2G and H). As expected from the growth rates and glucose tolerance, both male and female 138 βPHD3KO mice displayed similar insulin sensitivity to their βPHD3CON littermates ( Figure 2I  139 and J). Finally, no differences in islet size distribution ( Figure 2K) and β-cell mass ( Figure 2L

Loss of PHD3 improves insulin secretion at the onset of metabolic stress 153
We next examined whether PHD3 might play a more important role in regulating insulin 154 release during metabolic stress. Therefore, male animals were placed on high fat diet (HFD) 155 to induce obesity and metabolic stress (42). 156 Following 4 weeks HFD, Egln3 was moderately upregulated in βPHD3CON islets ( Figure 4A). 157 However, Egln3 levels remained suppressed in 4 weeks HFD βPHD3KO islets ( Figure 4A). 158 Glucose tolerance testing revealed significantly impaired glucose homeostasis in βPHD3KO 159 mice at 4 weeks but not at 72 hrs HFD ( Figure 4B and C), despite similar body weight gain 160 compared to βPHD3CON littermates ( Figure 4D). The 72 hour timepoint was used to 161 differentiate effects of early and prolonged fatty acid incorporation/utilization. As expected, 162 fasting blood glucose levels were elevated in βPHD3CON mice following 4 weeks HFD ( Figure  163 4C). There was no effect of Cre or flox'd alleles per se on metabolic phenotype following 4 164 weeks HFD, with Ins1 wt/wt ;Egln3 fl/fl , Ins1Cre +/-;Egln3 wt/wt and Ins1 wt/wt ;Egln3 wt/wt controls being 165 indistinguishable ( Figure 4E). IPGTT at 4 weeks HFD showed no difference in the serum 166 insulin levels between the βPHD3CON and βPHD3KO under fasting and glucose-stimulated 167 conditions ( Figure 4F). However, βPHD3KO mice mounted earlier and larger magnitude 168 insulin secretory responses to glucose bolus, as shown by the stimulation index ( Figure 4G). 169 Islets isolated from the same animals secreted significantly more insulin in glucose-stimulated 170 and Ex4-potentiated states ( Figure 4H), while insulin content was similar to βPHD3CON 171 littermates ( Figure 4I). Finally, 4 weeks HFD had no effect on glucose tolerance during OGTT 172 Thus, βPHD3KO mice are glucose-intolerant on HFD, show improved insulin secretion and 175 are able to release a greater fraction of their insulin granules (i.e. are more sensitized to 176 exocytosis). These data raise the possibility that nutrient-sensing and utilization might be 177 altered in βPHD3KO islets. 178

PHD3 maintains glycolysis and pyruvate management in β-cells 179
Given the reported roles of PHD3 in glycolysis, we wondered whether the changes in β-cell 180 function observed during the early phases of high fat feeding in βPHD3KO mice might be 181 associated with changes in glucose metabolism. We first looked at glycolytic fluxes using 14 C 182 glucose. While glucose oxidation was not altered at low or high glucose in islets from 4 weeks 183 HFD βPHD3KO mice ( Figure 5A), there was a small but significant decrease in 14 C content in 184 the aqueous phase, indicating a net reduction in tricarboxylic acid (TCA) cycle/other 185 metabolites derived from glycolysis ( Figure 5B). Notably, a 2-fold reduction in incorporation of 186 glucose into the lipid pool (i.e. glucose-driven lipogenesis) was also detected in 4 weeks HFD 187 βPHD3KO islets ( Figure 5C), suggestive of decreased oxidative pyruvate entry into the TCA 188 cycle and lipogenic acetyl-CoA (43). 189 To gain a higher resolution analysis of glucose fate, stable isotope-resolved tracing was 190 performed in βPHD3KO islets using 13 C6-[U]-glucose. The schematic in Figure 5D  Further analysis of steady-state lactate levels showed a significant increase in lactate 203 production in islets from HFD-fed βPHD3KO versus βPHD3CON mice ( Figure 5K). Together 204 with the m+2 → m+3 switch, this finding confirms initial measures with 14 C glucose indicating 205 reduced fueling of the TCA cycle by glycolysis ( Figure 5B). Furthermore, the tracing data 206 suggest that 4 weeks HFD βPHD3KO islets increase the reduction of pyruvate → lactate to 207 support continued glycolysis through regeneration of the cytosolic NAD + pool. While 208 expression of the "disallowed gene" Ldha (31, 32) tended to be increased, this was variable 209 and not significantly different between βPHD3CON and βPHD3KO islets. ( Figure 5L). 210 Together, these data suggest that metabolic stress induces defects in the management of 211 pyruvate fate in βPHD3KO islets, implying that insulin secretion in vitro must be maintained 212 and even amplified through mechanisms other than glycolysis. 213

PHD3 suppresses fatty acid use under metabolic stress 214
We hypothesized that βPHD3KO islets might switch to an alternative energy source to sustain 215 their function, namely beta oxidation of fatty acids, which are present in excess during HFD. 216 Moreover, in cancer cells PHD3 has been shown to increase activity of ACC2, which converts 217 acetyl-CoA → malonyl-CoA, the latter suppressing carnitine palmitoyltransferase I (CPT1), the 218 rate-limiting step in fatty acid oxidation (24, 44). Indicating a profound change in β-cell nutrient 219 preference, supplementation of culture medium with the fatty acid palmitate for 48-72 hrs 220 augmented glucose-stimulated and Exendin4-potentiated insulin secretion in 4 weeks HFD 221 βPHD3KO islets ( Figure 6A). By contrast, 4 weeks HFD βPHD3CON islets showed no 222 increase in glucose-stimulated insulin release following culture with palmitate ( Figure 6B), 223 confirming that the fatty acid was unlikely to induce lipotoxicity at the concentration and timing 224 used here. Interestingly, 48-72 hrs incubation with palmitate increased Exendin4-potentiated 225 insulin secretion in 4 weeks HFD βPHD3CON islets ( Figure 6B). 226 Providing evidence for increased CPT1 activity in 4 weeks HFD βPHD3KO islets, the CPT1a 227 inhibitor etomoxir was able to augment ATP/ADP responses to glucose in 4 weeks HFD 228 βPHD3KO relative to βPHD3CON islets ( Figure 6C), although mRNA levels of Cpt1a were 229 similar ( Figure 6D). In line with this finding, culture with low palmitate concentration decreased 230 glucose-stimulated Ca 2+ fluxes in 4 weeks HFD βPHD3KO but not in βPHD3CON islets 231 ( Figure 6E and F), presumably due to increased flux of fatty acid-derived acetyl-CoA into the 232 TCA cycle. While glucose-driven Ca 2+ fluxes were apparently normal in 4 weeks HFD 233 βPHD3KO islets, this was likely due to increased sensitivity of voltage-dependent Ca 2+ 234 channel to membrane depolarization, since responses to KCl were significantly elevated 235 ( Figure 6G and H). 236 To gain a higher resolution view of fatty acid fate, we incubated 4 weeks HFD βPHD3CON 237 and βPHD3KO islets with D31-palmitate, before measurement of intracellular D31-palmitate 238 concentration and 2H20 released from fatty acid oxidation. With this assay, the ratio of 2H20 239 to intracellular D31-palmitate provides a measure of fatty acid oxidation, whilst accounting for 240 any differences between tracer uptake/turnover. Confirming accuracy of the assay, 2H20/D31-241 palmitate values were robustly increased after 16 hrs versus 2 hrs incubation with tracer 242 ( Figure 6I).Notably, 2H20/D31-palmitate values were significantly higher in 4 weeks HFD 243 βPHD3KO versus βPHD3CON islets at the 16 hrs timepoint ( Figure 6I), indicative of higher 244 fatty acid oxidation rates. Uptake of tracer was similar in βPHD3KO versus βPHD3CON islets 245 ( Figure 6J). 246 Taken together, these data strongly suggest that PHD3 loss leads to alterations in fatty acid 247 utilization in islets. 248

Loss of PHD3 decreases dependency on glucose as a fuel source 249
We wondered whether increased fatty acid utilization in 4 weeks HFD βPHD3KO islets was 250 associated with a decreased dependency on glucose as a primary fuel source. Confirming a 251 switch away from glycolysis, glucose-stimulated ATP/ADP ratios were markedly decreased in 252 4 weeks HFD βPHD3KO islets ( Figure 6K and L), despite the apparent increases in insulin 253 secretion ( Figure 6A). Moreover, steady-state pyruvate levels were decreased in 4 weeks HFD 254 βPHD3KO islets ( Figure 6M). Lastly, glucose-stimulated insulin secretion (GSIS) was impaired 255 in SC βPHD3KO islets that were starved of glucose (3 mM) for 3 hrs prior to challenge ( Figure  256 6N), presumably due to dysregulated use of alternative fuel sources, which then inhibit critical 257 metabolic hubs in central carbon metabolism, such as pyruvate dehydrogenase. These data 258 further confirm the presence of defective pyruvate handling and suggest that βPHD3KO islets 259 alter pyruvate production and/or increase pyruvate → lactate conversion to maintain redox 260 balance during HFD. 261 Thus, following 4 weeks HFD, βPHD3KO islets become less reliant on glycolysis to fuel 262 ATP/ADP production, are able to sustain oxidative phosphorylation through fatty acid use, and 263 secrete more insulin when both glucose and fatty acids are present. 264

Regulated gene expression of ACC1 and ACC2 in β-cells 265
Previous studies have shown that PHD3 maintains glucose metabolism by hydroxylating and 266 activating ACC2 (encoded by Acacb), which inhibits CPT1 through generation of mitochondrial 267 malonyl-CoA, thus suppressing use of fatty acids via beta oxidation (45, 46). However, β-cells 268 are thought to predominantly express ACC1 (encoded by Acaca) (45, 46), which supplies 269 cytosolic malonyl-CoA to fatty acid synthase for de novo lipid biosynthesis rather than for beta 270 oxidation (43). Therefore, we sought to determine whether it was possible for PHD3 to act via 271 ACC2 in pancreatic β-cells. We re-examined the expression of ACACB in pancreatic β-cells 272 in multiple well-powered bulk islet and purified β-cell gene expression datasets (38, 47, 48). Thus, ACACB is present in β-cells, contains promoter regions regulated by β-cell-specific 286 transcription factors, but does not depend upon PHD3 for expression. These data are 287 consistent with a scenario whereby PHD3 hydroxylates ACC2 without influencing mRNA 288 expression. 289

PHD3 protects against insulin secretory failure during prolonged metabolic stress 290
Lastly, we sought to understand the phenotype of βPHD3KO mice when faced with continued 291 fatty acid/nutrient abundance. Glucose intolerance was still present in βPHD3KO mice 292 following 8 weeks on HFD ( Figure 7C), although less severe than at 4 weeks HFD, suggesting 293 that metabolic rewiring might in fact be protective against prolonged exposure to excess fatty 294 acids in vivo. As was the case at 4 weeks HFD, βPHD3KO mice showed similar insulin 295 sensitivity to βPHD3CON after 8 weeks HFD ( Figure 7D). In contrast to the IPGTT data, oral 296 glucose tolerance was preserved at this time point in βPHD3KO mice, suggesting an intact 297 incretin action ( Figure 7E). Furthermore, body composition of 8 weeks HFD βPHD3KO mice 298 was similar to βPHDCON ( Figure 7F). By this point, however, impaired glucose-stimulated 299 insulin secretion ( Figure 7G) was apparent in isolated βPHD3KO islets. This secretory deficit 300 could be rescued by application of Exendin4 to sensitize insulin granules to exocytosis ( Figure  301 7G and H), as expected from the OGTT results. In addition, the amplitude of glucose-302 stimulated Ca 2+ rises was significantly reduced in 8 weeks HFD βPHD3KO compared to 303 βPHD3CON islets ( Figure 7I and J). 304 Suggesting that profound defects in voltage-dependent Ca 2+ channels might also be present, 305 responses to the generic depolarizing stimulus KCl were markedly blunted in the same islets 306 ( Figure 7I and J). While apoptosis was increased in 8 weeks HFD βPHD3KO islets ( Figure 7K  307 and L), this did not reflect a (detectable) lipotoxic ER stress response, since Ddit3, Hspa5 and 308 Xbp1 ( Figure 7M) expression remained unchanged. Moreover, PCNA staining ( Figure 7N and 309 O) and α-cell/β-cell ratio ( Figure 7P and Q) were similar in 8 weeks HFD βPHD3CON and 310 βPHD3KO islets, suggesting that β-cells were unlikely to be less/more proliferative or adopting 311 α-cell features (or vice versa). Nonetheless, a profound 2-fold increase in β-cell mass was 312 observed in 8 weeks HFD βPHD3KO mice ( Figure 7R and S), with a significant increase in 313 the proportion of larger islets ( Figure 7T), implying that either: 1) apoptosis was restricted to 314 smaller/medium islets; or 2) changes in apoptosis/proliferation rate had not yet been able to 315 counter previous β-cell mass expansion. 316

Loss of PHD3 is not associated with changes in HIF stabilization 317
Previous studies have shown that PHD3 is highly regulated at the transcriptomic level by 318 hypoxia (10), and in line with this, we also found that Egln3 levels in WT islets were increased In the present study, we show that the alpha-ketoglutarate-dependent PHD3 maintains β-cell 337 glucose sensing under states of metabolic stress associated with fatty acid abundance. Our 338 data suggest that PHD3 is required for ensuring that acetyl-CoA derived from glycolysis 339 preferentially feeds the TCA cycle, linking blood glucose levels with ATP/ADP generation, β- the levels of ACACB, which encodes ACC2, were found to be similar to the β-cell transcription 360 factor HNF1A, albeit lower than those of ACACA. We thus propose that loss of PHD3 might 361 plausibly lead to suppression of ACC2 activity, which becomes apparent during HFD when its 362 substrate is present in abundance. Alternatively, PHD3 might hydroxylate and activate ACC1, 363 leading to regulation of CPT1 by malonyl-CoA when fatty acids are supplied in excess, as 364 suggested by glucose oxidation experiments. In both cases, identifying the PHD3 365 hydroxylation sites involved will be critical. However, assigning hydroxylation targets using 366 mass spectrometry is currently controversial: mis-alignment of hydroxylation is frequently 367 associated with the presence of residues in the tryptic fragment that can be artefactually 368 oxidized (44, 50). Thus, studies using animals lacking PHD3 and ACC1/ACC2 in β-cells, or 369 alternatively the use of (relatively) specific inhibitors, would be required to definitively link the 370 carboxylase with the phenotype here. 371 As normal chow contains a low proportion of calories from fat, metabolic stress was 372 needed to reveal the full in vitro and in vivo phenotype associated with PHD3 loss. These data 373 also support an effect of PHD3 on ACC1/2 and CPT1, since without acyl-CoA derived from 374 exogenous fatty acids, glucose would still constitute the primary fuel source and regulator of 375 insulin release. The lack of phenotype under normal diet is unlikely to reflect the age of the 376 animals, since even at 20 weeks of age, glucose intolerance was still not present in βPHD3KO 377 mice. Of interest, the severity of the βPHD3KO in vivo phenotype was milder at 8 weeks versus 378 4 weeks HFD feeding, despite the presence of impaired glucose-dependent β-cell function by 379 this timepoint. These observations suggest that, by 8 weeks HFD, compensatory protective 380 mechanisms may become upregulated as a consequence of the metabolic re-wiring in β-cells. 381 It will be necessary in the future to investigate the mechanistic/phenotypic changes occurring 382 during even longer duration HFD feeding (e.g. 12-20 wks). It will also be interesting to 383 understand how PHD3 activity changes in other models of metabolic stress, such as db/db 384 and ob/ob mice. 385 Suggesting that the phenotype associated with PHD3 loss was not due to changes in 386 HIF signaling, no differences in the gene expression of HIF1 targets could be detected in 387 βPHD3KO versus βPHD3CON islets. Indeed, PHD2 is the major hydroxylase that regulates 388 HIF1α stability (11, 12), with no changes in activity of the transcription factor following PHD3 389 loss (11, 12, 51). Thus, it is perhaps unsurprising that there is a lack of HIF1 transcriptional 390 signature in βPHD3KO islets, in agreement with previous studies in other tissues (51, 52). In 391 addition, glucose-stimulated Ca 2+ fluxes, a sensitive readout of changes in oxygen-dependent 392 regulation (49), were unaffected during hypoxia in βPHD3KO islets. While there was a trend 393 toward increased Ldha expression in HFD βPHD3KO islets, this was just a fraction of that 394 previously reported in hypoxic rodent islets (53). Nonetheless, we cannot completely exclude 395 HIF-dependent effects, and as such, studies should either be repeated on a HIF1-and HIF2-396 null background (i.e. a quadruple transgenic) or using (moderately) specific chemical 397

inhibitors. 398
We acknowledge a number of limitations with the present studies. Firstly, work-up was 399 limited to rodents and it will be important to confirm whether results translate to human islets 400 or not. While our attempts at silencing PHD3 using EGLN3 shRNA were unsuccessful, studies 401 using (relatively) specific PHD3 inhibitors are warranted. Secondly, interactions between 402 PHD3 and ACC2 are inferred from our metabolic work up and known biochemistry. Identifying 403 hydroxylation sites and creating corresponding ACC1/2 mutants is needed, but current mass 404 spectrometry analysis is challenging due to the assignment of false positives, as mentioned 405 above. Thirdly, we focused our studies on 4 and 8 weeks HFD and it is unclear whether the 406 switch toward increased fatty acid utilization might be maladaptive or protective in βPHD3KO 407 mice during longer periods of HFD feeding. Fourthly, HFD studies were restricted to male 408 animals and further studies should be extended to female animals. While sex differences in 409 phenotype were not observed under standard diet, we cannot exclude a sexually dimorphic 410 effect of HFD. In summary, PHD3 possesses a conserved role in gating nutrient preference 411 toward glucose and glycolysis during both cell transformation (24) and metabolic stress (as 412 shown here). It will be interesting to now study whether similar effects of PHD3 are present in 413 other cell types involved in glucose-sensing (for example, pancreatic alpha cells, hypothalamic 414 neurons). 415

Experimental design 417
No data were excluded unless the cells displayed a non-physiological state (i.e. impaired 418 viability). All individual data points are reported. The measurement unit is animal or batch of 419 islets, with experiments replicated independently. Animals and islets were randomly allocated 420 to treatment groups to ensure that all states were represented in the different experiment arms. 421

Mouse models 422
β-cell-specific PHD3 (βPHD3KO) knockout mice were generated using the Cre-LoxP system. 423 until 18-20 weeks of age. Animals were maintained in a specific pathogen-free facility, with 438 free access to food and water. 439

Intraperitoneal and oral glucose tolerance testing 440
Mice were fasted for 4-6 hrs, before intraperitoneal injection of glucose. Animals on SC 441 received 2 g/kg body weight glucose, whereas those on HFD received a lower dose of 1 g/kg 442 body weight. In our hands, this allows measurement of blood glucose concentration without 443 the need to dilute samples and decreases adverse reactions associated with profound 444 hyperglycemia. Blood samples for glucose measurement were taken from the tail vein at 0, 445 15, 30, 60, 90 and 120 min post-challenge. Glucose was measured using a Contour XT 446 glucometer (Bayer, Germany). For mice on SC, intraperitoneal glucose tolerance testing 447 (IPGTT) was performed every 2-4 weeks, between 8-20 weeks of age. HFD-fed mice 448 underwent IPGTT following 72 hrs, 4 and 8 weeks of HFD. Oral glucose tolerance testing 449 (OGTT) was performed as for IPGTT, except that glucose was delivered using an oral gavage 450 tube (2 g/kg and 1 g/kg body weight in SC-fed and HFD-fed mice, respectively) . 451

Serum insulin 452
Blood samples were collected following intraperitoneal glucose injection (1 g/kg body weight).

Gene expression 474
Trizol purification was used for mRNA extraction, while cDNA was synthesized by reverse 475 transcription. Gene expression was detected by real time PCR (qPCR), using PowerUp SYBR 476 Green Master Mix (Thermofisher Scientific) and quantification was based on the 2 -ΔΔCt method, 477 expressed as fold-change in gene expression. The sequence of the forward and reverse 478 primers used in the study can be found in Supplementary Table 1.

Visualization of transcriptomic datasets 588
Details of the RNA-seq and ChIP-seq experiments, as well as human islet donors, are 589 Measurements were performed on discrete samples unless otherwise stated. Data normality 597 was assessed using D'Agostino-Person test. All analyses were conducted using GraphPad 598 Prism software. Pairwise comparisons were made using Student's two-tailed unpaired or 599 paired t-test. Multiple interactions were determined using one-way ANOVA or two-way 600 ANOVA, adjusted for repeated measures where relevant. Pairwise post-hoc testing was 601 performed using Sidak's test, or Tukey's test where more than two groups were considered. 602 Where a highly significant interaction was detected using two-way ANOVA, but post-hoc 603 testing was just above P = 0.05, multiple comparisons were accounted for using the false 604 discovery rate followed by the two-stage linear step-up method of Benjamini, Krieger and 605 Yekutieli. For non-parametric multiple comparison, Kruskal-Wallis test was used followed by 606 Dunn's post hoc test. Degrees of freedom were accounted for during all post-hoc testing. A P 607 value less than 0.05 was considered significant. 608

Data availability 609
The datasets generated and/or analyzed during the current study are available from the 610 corresponding author upon reasonable request. 611