The E3 ubiquitin ligase parkin is dispensable for metabolic homeostasis in murine pancreatic β cells and adipocytes

The E3 ubiquitin ligase parkin is a critical regulator of mitophagy and has been identified as a susceptibility gene for type 2 diabetes (T2D). However, its role in metabolically active tissues that precipitate T2D development is unknown. Specifically, pancreatic β cells and adipocytes both rely heavily on mitochondrial function in the regulation of optimal glycemic control to prevent T2D, but parkin's role in preserving quality control of β cell or adipocyte mitochondria is unclear. Although parkin has been reported previously to control mitophagy, here we show that, surprisingly, parkin is dispensable for glucose homeostasis in both β cells and adipocytes during diet-induced insulin resistance in mice. We observed that insulin secretion, β cell formation, and islet architecture were preserved in parkin-deficient β cells and islets, suggesting that parkin is not necessary for control of β cell function and islet compensation for diet-induced obesity. Although transient parkin deficiency mildly impaired mitochondrial turnover in β cell lines, parkin deletion in primary β cells yielded no deficits in mitochondrial clearance. In adipocyte-specific deletion models, lipid uptake and β-oxidation were increased in cultured cells, whereas adipose tissue morphology, glucose homeostasis, and beige-to-white adipocyte transition were unaffected in vivo. In key metabolic tissues where mitochondrial dysfunction has been implicated in T2D development, our experiments unexpectedly revealed that parkin is not an essential regulator of glucose tolerance, whole-body energy metabolism, or mitochondrial quality control. These findings highlight that parkin-independent processes maintain β cell and adipocyte mitochondrial quality control in diet-induced obesity.

Mitochondrial health is predicated on an intricate balance between biogenesis of new functional mitochondria and turnover of old, dysfunctional mitochondria. Selective turnover of dysfunctional mitochondria via the autophagy machinery, also known as mitophagy, is an essential quality control mechanism ensuring mitochondrial health. To date, the most comprehensively studied pathway of mitophagy is initiated by the E3 ubiquitin ligase parkin, encoded by the Prkn gene (1). Parkin targets damaged mitochondria for turnover following its recruitment to the mitochondrial surface by PTEN-induced kinase 1 (PINK1). Following ubiquitination of its substrate proteins, including Mitofusin 1 and 2 (MFN1/MFN2) and VDAC1, parkin initiates the recruitment of autophagy receptors necessary to ferry mitochondria within autophagosomes to the lysosome for degradation (2,3). Mutations or deficiency of parkin is associated with both heritable and sporadic forms of Parkinson's disease, and the incidence of type 2 diabetes (T2D) 5 is higher in the Parkinson's population, suggesting a connection between parkin and T2D in humans (4,5). Moreover, there is genetic evidence of associations between parkin and T2D (6 -11).
T2D is a chronic multisystem disease manifested by the combination of peripheral insulin resistance and insufficient insulin secretion. Pancreatic ␤ cells and adipocytes are two vital contributors to the development of T2D. Indeed, mitochondrial structure, morphology, and function are impaired in ␤ cells and adipocytes in T2D (12)(13)(14), which is suggestive of impaired mitophagy. Mitophagy is critical for ␤ cell function (15,16); however, the role of parkin-dependent mitophagy in ␤ cells is unclear because of conflicting studies demonstrating both protective and disruptive roles for parkin deficiency in ␤ cell function (7,(17)(18)(19). On the other hand, whole-body Prkn-null mice are protected from diet-induced obesity (DIO) (6), and more recently, these animals have been shown to have prolonged maintenance of metabolically beneficial beige/BRITE adipose . The JDRF career development award to S. A. S. is partly supported by the Danish Diabetes Academy, which is supported by the Novo Nordisk Foundation. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Figs. S1-S7 and Tables S1 and S2. 1 Both authors contributed equally to this work. 2 Supported by NIDDK, NIH Summer Undergraduate Research Fellowship R25 DK088752. 3 To whom correspondence may be addressed. Tel.: 734-647-4880; Fax: 734-232-8175; E-mail: macdouga@med.umich.edu. 4 To whom correspondence may be addressed. Tel.: 734-232-8150; Fax: 734-232-8162; E-mail: ssol@med.umich.edu.
tissue following withdrawal of ␤3-adrenergic activation (20 -22). This beige-to-white adipocyte transition is highly dependent on clearance of mitochondria via mitophagy and is impaired in global Prkn-null mice (20). It is well-established that there are diverse roles for parkin in metabolic tissues, but precise tissue-specific functions of parkin-mediated mitophagy in the development of T2D have not been elucidated.
Here we report that parkin is not required for pancreatic development or ␤ cell function, that responses to DIO are unchanged following ␤ cell parkin deficiency, and that parkin is dispensable for mitochondrial turnover following damage in ␤ cells. We further demonstrate that loss of parkin in adipocytes does not modulate whole-body glucose metabolism, adipocyte morphology, or mitochondrial mass. We identify a role for parkin in ␤-oxidation of fatty acids within adipocytes but not in preadipocyte differentiation. Finally, we determine that parkin is not required within adipocytes for beige-to-white adipocyte transition following cessation of cold exposure or ␤3-adrenergic stimulation.

Parkin deficiency in pancreatic endocrine cells does not affect glucose tolerance
The importance of mitochondria and mitochondrial turnover (mitophagy) in pancreatic endocrine cells cannot be understated (15,16,(23)(24)(25)(26). However, the role of the key initiator of mitophagy, parkin, specifically in pancreatic endocrine cells has yet to be fully understood. Indeed, there are conflicting reports in the literature regarding parkin's contribution to ␤ cell function (7,(17)(18)(19). To this end, we first investigated the role of parkin in pancreatic cells by utilizing the Prkn FL/FL floxed mouse (hereafter known as Parkin flox ) crossed to the Pdx1-Cre mouse (27) to elucidate the role of parkin in pancreatic islet function (Parkin flox ; Pdx1-Cre; hereafter called Panc-Parkin KO ). Because of the expression of Pdx1 in pancreatic development, this model allows investigation into parkin's involvement during development of the endocrine pancreas as well as in mature islets after birth (28). We investigated the role of parkin both at baseline as well as after high-fat diet (HFD) feeding to induce obesity as a diabetogenic stressor. Panc-Parkin KO islets exhibited loss of parkin protein expression compared with Pdx1-Cre-only controls ( Fig. 1A) but maintained normal glucose tolerance at 10 weeks of age (Fig. 1B). We observed no fasting hypoglycemia (Fig. 1B), suggesting no defects in ␣ cell function. We also observed no steatorrhea, suggesting that parkin-deficient mice do not develop overt pancreatic exocrine function (data not shown).
Next we sought to understand the role of parkin in response to obesity-related metabolic stress. Pdx1-Cre control and Panc-Parkin KO mice were fed an HFD at weaning (Fig. 1C) and gained weight similarly overall, with the exception of a subtle but significant decrease in body weight in 19-week-old Panc-Parkin KO mice (Fig. 1C). Glucose tolerance was again unchanged at this age ( Fig.  1D), nor was any significant difference observed in glucose tolerance pre-HFD initiation or at 4 and 8 weeks of HFD feeding (Fig.  S1, A-C). Histological analysis of pancreatic sections revealed normal islet architecture in Panc-Parkin KO mice, with no differences in ␤ or ␣ cell distribution by insulin and glucagon immunostaining, respectively (Fig. 1E). Altogether, these data suggest that parkin is dispensable for islet formation and glucose homeostasis at baseline and following DIO.

Parkin is dispensable for pancreatic ␤ cell function at baseline and following DIO
Because of recent concerns regarding the study of pancreatic islet growth in Pdx1-Cre mice related to expression of the human growth hormone minigene (29), we wanted to further A, representative immunoblot of parkin protein expression in isolated pancreatic islets from either Pdx1-Cre or Panc-Parkin KO mice. B, blood glucose concentrations during an IPGTT of Pdx1-Cre (blue circles) or Panc-Parkin KO (orange squares) mice on a normal chow diet at 10 weeks of age (n ϭ 5/group). C, weights over duration of HFD feeding in Pdx1-Cre or Panc-Parkin KO mice (n ϭ 4 -5/group). D, blood glucose concentrations during an IPGTT of Pdx1-Cre (blue circles) or Panc-Parkin KO (orange squares) mice after being fed HFD for 12 weeks (n ϭ 4 -5/group). E, representative immunofluorescence images of pancreatic sections from Pdx1-Cre or Panc-Parkin KO mice stained for insulin (green), glucagon (red), and DAPI (blue). *, p Ͻ 0.05; two-way ANOVA with Sidak's multiple comparisons post-test.

Parkin function in ␤ cells and adipocytes
confirm the role of parkin specifically in pancreatic ␤ cells utilizing Parkin flox mice crossed to Ins1-Cre knock-in mice (30) (hereafter called ␤-Parkin KO ). ␤-Parkin KO mice also demonstrated loss of parkin protein in isolated islets ( Fig. 2A). No significant differences were observed in body weight or glucose tolerance between Ins1-Cre-alone or Parkin flox , Parkin flox/ϩ , or Parkin ϩ/ϩ -alone control mice (Fig. S2, A-C); thus, all studies utilized a mixture of control (Ctrl) animals. Similar to findings following pancreas-specific parkin loss of function ( Fig. 1), glucose tolerance in ␤-Parkin KO mice was indistinguishable from controls at 8 weeks of age (Fig. 2B).
To further elucidate the role of parkin in ␤-cell function, GSIS was assessed in vivo to investigate whether ␤ cell secretory function is impacted by loss of parkin. Interestingly, 9-week-old ␤-Parkin KO mice exhibited higher insulin release 3 min after a glucose challenge compared with controls (Fig. 2C). However, this was not accompanied by changes in total pancreatic insulin content (Fig. 2D), glucose tolerance (Fig. 2B), or islet morphology (Fig. 2E). These data indicate that ␤-Parkin KO mice could have the capacity for enhanced insulin secretion to potentially drive improved glucose clearance, but no changes in glucose clearance were observed (Fig. 2B). Taken together, these data again suggest that loss of parkin is not detrimental to ␤ cell function or whole-body glucose homeostasis.
To determine a role of parkin during metabolic stress, ␤-Parkin KO mice and littermate controls were placed on an HFD at weaning and monitored for 4 months. Both male and female Ctrl and ␤-Parkin KO mice gained weight similarly throughout the study (Fig. 2F), and, as seen previously (Fig. 1D), no difference in glucose tolerance was observed between geno-Figure 2. Parkin is not required for pancreatic ␤ cell function, either at baseline or after DIO. A, representative immunoblot of parkin protein expression in isolated islets from Ins1-Cre or ␤-Parkin KO mice. B, blood glucose concentrations during an IPGTT of Ins1-Cre (blue diamonds) or ␤-Parkin KO (orange triangles) mice at 8 weeks of age (n ϭ 4 Ins1-Cre and 7 ␤-Parkin KO mice). C, plasma insulin concentrations at baseline (0 min) and 3 min after a 3 mg/kg glucose bolus in Ins1-Cre (blue diamonds) and ␤-Parkin KO (orange triangles) mice at 9 weeks of age (n ϭ 4 Ins1-Cre and 7 ␤-Parkin KO mice). D, total pancreatic insulin content from Ins1-Cre (blue column) and ␤-Parkin KO (orange column) pancreata at 9 weeks of age (n ϭ 4 Ins1-Cre and 7 ␤-Parkin KO mice). E, representative immunofluorescence images of pancreatic sections from Ins1-Cre or ␤-Parkin KO mice stained for insulin (green), glucagon (red), and DAPI (blue). F, weights of male and female Ins1-Cre or ␤-Parkin KO mice following HFD feeding (males, n ϭ 7 Ins1-Cre and 10 ␤-Parkin KO mice; females, n ϭ 7 Ins1-Cre and 12 ␤-Parkin KO mice). G and H, blood glucose concentrations during an IPGTT of Ins1-Cre (blue diamonds) and ␤-Parkin KO (orange triangles) mice after 4 weeks (G) or 16 weeks (H) of HFD feeding (n ϭ 15 Ins1-Cre and 23 ␤-Parkin KO mice). I, representative immunofluorescence images of pancreatic sections from 20-week HFD-fed Ins1-Cre or ␤-Parkin KO mice stained for insulin (green), glucagon (red), and DAPI (blue). *, p Ͻ 0.05; two-way ANOVA with Sidak's multiple comparisons post-test.
types throughout the HFD study (Fig. 2, G and H, and Fig. S1, D and E). Additionally, islet morphology was unchanged between Ctrl and ␤-Parkin KO mice. These data confirm that loss of parkin is dispensable for ␤ cell adaptation to DIO and that ␤ cells deficient in parkin are fully capable of regulating whole-body glucose homeostasis.

Parkin has mild effects on mitochondrial turnover in pancreatic ␤ cells
Although it is evident that parkin is dispensable for ␤ cell function (Figs. 1 and 2), parkin is known to be a pivotal node in mitochondrial turnover in a number of other cell types (31)(32)(33). Therefore, we wanted to examine whether parkin deficiency affects mitochondrial turnover in ␤ cells. Utilizing parkin-specific siRNA, we transiently knocked down parkin in MIN6 ␤ cells. Following an ϳ40 -50% reduction in parkin protein levels ( Fig. 3A), we observed that expression of the outer mitochondrial membrane protein MFN2 was decreased, whereas another outer membrane protein, VDAC1, was not similarly affected. We next examined whether the rate of mitochondrial turnover was impacted by parkin loss following treatment with the mitochondrial uncoupler carbonyl cyanide m-chlorophenylhydra-

Parkin function in ␤ cells and adipocytes
zone (CCCP), which is known to dissipate mitochondrial membrane potential and initiate clearance via mitophagy (34). CCCP treatment caused a time-dependent decrease in MFN2 and VDAC1 in control nontargeting (NT) siRNA-treated MIN6 ␤ cells (Fig. 3, B and C); however, acute parkin deficiency significantly slowed the rate of turnover. These data suggest that acute loss of parkin does affect mitophagy in ␤ cells following robust mitochondrial damage (Fig. 3, B and C), but acute parkin deficiency does not appear to impact ␤ cell function, as GSIS continues to be unaffected (Fig. 3D). Similarly, cellular stress responses are unaffected, as reactive oxygen species (ROS) generation is not different following acute loss of parkin (Fig. S3A).
To further investigate the role of parkin in ␤ cell mitochondrial turnover, isolated islets from control or ␤-Parkin KO mice were treated ex vivo with the ionophore valinomycin to again dissipate mitochondrial membrane potential and induce mitochondrial clearance via mitophagy (34). Deletion of parkin in vivo had no effect on expression of outer mitochondrial membrane proteins at baseline (Fig. 3E). Surprisingly, we observed no overt effect of parkin on the rate of mitochondrial turnover after valinomycin treatment in primary islets (Fig. 3F). We also observed no significant differences in bulk autophagy machinery, as measured by LC3 and p62 protein levels in Ctrl or ␤-Parkin KO islets ex vivo (Fig. S3, B and C). Taken together, these results highlight that parkin is not required for mitochondrial turnover in ␤ cells in vivo and has only a small effect on turnover after transient loss of function. These findings could suggest a novel and potentially important role for parkin-independent mitophagy (35,36), which may maintain appropriate ␤ cell mitochondrial quality control in the absence of parkin.

Body weight, adiposity, and glucose tolerance are not affected by adipose-specific loss of parkin
The roles of parkin appear to be minimal in ␤ cell responses to excess metabolic demand; thus, we next investigated whether its role in adipose tissue, which also plays a causative role in T2D, elicited more of a phenotype. Parkin has been described as a regulator of fat uptake, as global parkin-null mice are resistant to the weight gain, hepatic steatosis, and insulin resistance caused by feeding an HFD (6). To investigate cellautonomous roles of parkin regulation in lipid metabolism in adipocytes, we generated mice lacking parkin in adipose tissue by crossing Parkin flox with Adiponectin-Cre mice to generate Prkn FL/FL ;Adiponectin-Cre/ϩ mice (AD-Parkin KO ). Adiponectin-Cre is a well-established model to delete floxed genes selectively and efficiently in adipocytes with minimal off-target effects (37). In contrast to the findings by Kim et al. (6) with global parkin deficiency, AD-Parkin KO and Parkin flox littermate controls did not have differences in weight gain over the course of 12 weeks of feeding a high-fat (Fig. 4A) or normal chow diet (data not shown). Adipocyte-specific deletion of parkin was confirmed by genotyping of DNA (Fig. S4, A and B), immunoblot of protein extracts from adipose tissue (Fig. S4C), and A, body weights of Parkin flox and AD-Parkin KO mice over the course of 12 weeks of HFD feeding (n ϭ 7-9 animals/group). B, relative expression of Prkn mRNA in adipocytes (AD) and the stromal vascular fraction (SVF) isolated from iWAT of Parkin flox and AD-Parkin KO mice after 12 weeks of HFD feeding (n ϭ 2 animals/group). C, body composition measured by NMR spectroscopy after 12 weeks of HFD feeding (n ϭ 5 animals/group). D, representative histological images of the iWAT, gonadal WAT (gWAT), brown adipose tissue (BAT), and liver after 12 weeks of HFD feeding. E, glucose tolerance after 12 weeks of HFD feeding. Mice were fasted for 16 h and then injected with 1 mg/kg glucose intraperitoneally. Shown are blood glucose concentrations during an IPGTT in Parkin flox and AD-Parkin KO mice at the indicated time points (n ϭ 5 animals/group). F, blood glucose concentrations during random ad libitum feeding or after fasting (16-h food restriction) following 12 weeks of HFD (n ϭ 5 animals/group). *, p Ͻ 0.05; Student's unpaired t test.

Parkin function in ␤ cells and adipocytes
expression of Prkn mRNA (Fig. 4B). Body composition measured by NMR spectroscopy did not significantly differ between Parkin flox and AD-Parkin KO mice (Fig. 4C). Individual tissue weights from Parkin flox and AD-Parkin KO mice were also relatively similar after 12 weeks of HFD feeding (Fig. S4D). Histological analysis of various tissues did not reveal gross abnormalities in cell size or morphology between AD-Parkin KO mice and control littermates (Fig. 4D).
Next we asked whether AD-Parkin KO mice had metabolic alterations relative to the Parkin flox controls despite the lack of obvious changes in body weight, tissue weight, or tissue morphology. Neither glucose tolerance nor insulin sensitivity were significantly changed in AD-Parkin KO mice compared with the Parkin flox controls following HFD feeding (Figs. 4E and S4E). Fasting and random-fed blood glucose concentrations were also similar between experimental groups (Figs. 4F and S4F). We considered whether glucose intolerance might be masked by compensatory release of insulin; however, the concentration of insulin in the serum of Parkin flox and AD-Parkin KO mice was similar in both fasting and fed states (Fig. S4G). Circulating glycerol concentrations in the serum of Parkin flox and AD-Parkin KO mice were also unaffected by parkin deletion (Fig.  S4H), suggesting that parkin is not required for regulation of lipolysis. As parkin is viewed to be a critical regulator of mitophagy and to signal the clearance of damaged mitochondria (34), we also performed transmission EM to observe the morphology and integrity of mitochondria in brown adipose tissue of Parkin flox and AD-Parkin KO mice. The size, number, and structure of the mitochondria appeared to be unaffected by parkin deletion (Fig. S4I), suggesting that other pathways may compensate to regulate mitochondrial integrity in the absence of parkin expression (35,36). Alternatively, other cell types within the adipose tissue that retain parkin expression may signal to adipocytes through unknown mechanisms to maintain mitochondrial homeostasis. Finally, we subjected the experimental mice to a variety of metabolic analyses using comprehensive lab animal monitoring system (CLAMS). No significant differences in food intake, physical activity, oxygen consumption, carbon dioxide production, respiratory exchange ratio, or glucose oxidation were observed between the experimental groups (Fig. S5). Energy expenditure and fat oxidation were also unchanged (data not shown). Together, these data indicate that adipose-specific deletion of parkin does not affect global adiposity, glucose tolerance, or metabolic homeostasis in mice.

Parkin is not required for normal adipocyte differentiation but does play a role in adipocyte ␤-oxidation
To further investigate the molecular function of parkin in adipocytes, we isolated primary ear mesenchymal stem cells (eMSCs) from Parkin flox mice and subjected the cells to a variety of molecular and metabolic analyses. Parkin flox eMSCs were infected with an adenovirus expressing either GFP as a negative control (Ad-GFP) or Cre recombinase to induce parkin deletion (Ad-Cre). Recombination of the floxed allele was confirmed by PCR using primers flanking the loxP sites (Fig. S6A). These cells were then differentiated into mature adipocytes using standard adipogenic stimuli (insulin, dexamethasone, 3-isobutyl-1-methylxanthine (IBMX), and rosiglitazone). The ability of precursors to differentiate and the morphology of the mature adipocytes were not affected by parkin deletion, as observed by phase-contrast microscopy and Oil Red O staining (Fig. 5A).
Next we analyzed the ability of eMSC adipocytes to metabolize fatty acids and found that ␤-oxidation of [ 3 H]-labeled palmitic acid (Fig. 5B) or [ 3 H]oleic acid was significantly increased in Parkin flox -Ad-Cre adipocytes compared with Parkin flox -Ad-GFP controls (Fig. S6B). Etomoxir, a selective inhibitor of the mitochondrial fatty acid transporter carnitine palmitoyltransferase 1␣ (CPT1␣), blocked ␤-oxidation in both Parkin flox -Ad-GFP and Parkin flox -Ad-Cre adipocytes to a similar extent (Figs. 5B and S6B). Fatty acid uptake into eMSC adipocytes was also significantly increased in the absence of parkin (Fig. S6C). However, we did not observe differential ␤-oxidation when adipocytes were incubated with medium-chain [ 3 H]-labeled octanoic acid (Fig. S6D) in the presence or absence of etomoxir. These data indicate that the increase in ␤-oxidation is specific to long-chain fatty acids and depends on the activity of CPT1␣, which facilitates transport of long-chain fatty acids across the outer mitochondrial membrane. Indeed, we observed higher CPT1␣ protein and mRNA levels in Parkin flox eMSCs treated with Ad-Cre (Fig. 5, C and D). Interestingly, adiponectin protein was reduced in parkin-deficient cultured adipocytes (Fig.  5C), and expression of mRNA for the adipocyte markers AdipoQ and Fabp4 was reduced significantly in Parkin flox -Ad-Cre adipocytes (Fig. 5D) despite the lack of morphological changes in the cells (Fig. 5A). Expression of oxidative phosphorylation complex proteins was also slightly reduced in Parkin flox -Ad-Cre adipocytes (Fig. 5C); however, parkin deletion did not significantly alter expression of the mitochondrial or regulatory genes Cpt1␣, Cox1, and Pgc1␣ (Fig. 5D) or the total number of mitochondria (Fig. 5E). Protein expression of FABP4, MFN2, or VDAC1 were also unchanged (data not shown). In mature adipocytes, parkin deletion did not affect lipolytic activity, either in the basal state or when induced with forskolin, or insulin-stimulated glucose uptake (Fig. S6, E and F). We observed no differences in the generation of cellular ROS in Parkin flox -Ad-GFP and Parkin flox -Ad-Cre adipocytes (Fig. S6G). We also measured the levels of the autophagic proteins LC3 and p62 in Parkin flox -Ad-GFP and Parkin flox -Ad-Cre adipocytes and found no change in LC3 expression but an increase in p62 levels following parkin deletion (Fig. S6H). Together, these data demonstrate that parkin may influence specific aspects of lipid metabolism in cultured eMSC adipocytes; however, these changes are not significant enough to induce phenotypic changes in mice lacking parkin expression in adipose tissue.

Formation and reversion of beige adipose tissue is not dependent on parkin expression
Autophagy and mitophagy are essential for the maintenance of beige adipocytes as induced by cold exposure or ␤3-adrenergic stimulation (20,21). Lu et al. (20) recently reported that, although parkin is not required for development of beige adipose tissue, the beige-to-white adipocyte transition of inguinal white adipose tissue (iWAT) upon withdrawal of beige-inducing stimuli is impaired in global Prkn-null mice. Thus, we investigated whether parkin deletion in adipocytes impaired either

Parkin function in ␤ cells and adipocytes
the ability of iWAT to acquire beige characteristics or the ability of beige fat to revert back to WAT after cessation of cold exposure or ␤3-adrenergic stimulation.
To address these questions, we treated Parkin flox and AD-Parkin KO mice with the ␤3-adrenergic agonist CL-316,243 for 1 week to induce beige fat formation and then let the animals recover without drug administration for 15 days. The body weights and tissue weights did not differ between Parkin flox and AD-Parkin KO mice following CL-316,243 administration and recovery (Fig. 6, A and B). In contrast to the reported findings with the global parkin-null mice (20), differences were not observed in beige-to-white adipocyte transition upon cessation of CL-316,243 treatment, as evidenced by histological analysis (Fig. 6C) and expression of UCP1 and oxidative phosphorylation complex proteins (Fig. 6D). These changes were not dependent on diet, as we observed similar phenotypes in Parkin flox and AD-Parkin KO mice fed an HFD (Fig. S7, A-C). Furthermore, we did not observe any differences in the circulating levels of leptin or adiponectin in the serum from Parkin flox and AD-Parkin KO mice (Fig. S7, D and E).
To determine whether these phenotypes were dependent on the type of beige fat-inducing stimuli, we also subjected a group of female mice to cold exposure (6°C for 7 days), followed by 15 days of recovery at room temperature. Again, no significant differences in the capacity of iWAT to beige or transition back to white adipocytes were observed with parkin deletion (Fig. 7,  A-E). These findings demonstrate that adipocyte-specific deletion of parkin is insufficient to affect the formation or reversion of beige adipose tissue and suggest that other cell types contrib-ute to inhibition of the beige-to-white adipocyte transition observed in global parkin-null mice. These data indicate that loss of parkin in adipocytes does not affect adipose tissue morphology, expression of metabolic proteins, or maintenance of beige adipocytes following cold exposure or ␤3-adrenergic stimulation.

Discussion
Mitochondrial function and homeostasis are critical to maintain normal cellular activities. Disruption of mitochondrial quality control is implicated in numerous disease states, including obesity and ␤ cell dysfunction in T2D (12)(13)(14). Despite a large body of evidence identifying parkin as a critical regulator of mitophagy, we did not observe mitochondrial dysfunction in mice with pancreatic-, ␤ cell-, or adipocyte-specific parkin deletion or any phenotypes affecting glucose homeostasis or metabolic health. Our findings suggest that parkin is largely dispensable for adipose and pancreatic islet/␤ cell function and whole-body glucose homeostasis under a variety of metabolic conditions.
Parkin is considered to be a master regulator of mitophagy, and mitochondria and mitochondrial turnover are essential for proper cellular function, especially in pancreatic ␤ cells (12,15,16,24). Therefore, it was surprising that loss of a key mitochondrial quality control protein (i.e. parkin) elicited little to no phenotype. The role of parkin in ␤ cells has been inconclusive to date, with studies showing that loss of parkin results in impaired insulin release and production as well as increased susceptibility to streptozotocin-induced diabetes (7, 19) but

Parkin function in ␤ cells and adipocytes
also that overexpression or activation of parkin-dependent pathways results in aberrant ␤ cell function (17,18). Although studies in other systems have described the importance of parkin in the initiation of mitophagy, these studies were primarily performed in ex vivo cell-based systems following parkin overexpression and severe mitochondrial damage (34,36). The role of parkin in physiologically relevant contexts of mitophagy in vivo is still not well-developed. Our study demonstrates that, in the context of obesity caused by HFD consumption, parkin deficiency does not lead to ␤ cell failure. This could indicate that the stress of overnutrition does not exceed the capacity of ␤ cells to adapt to increased mitochondrial metabolic demand or that parkin may have a redundant role in mitochondrial turnover with other pathways. Indeed, here we identify that mitochondrial turnover remains largely intact following loss of parkin, indicating the likelihood of compensatory parkin-independent mechanisms.
Expression of parkin and its upstream activating kinase PINK1 increases during adipocyte differentiation and is also increased in white adipose tissue of mice fed an HFD relative to normal chow-fed controls (38,39). This suggests a role of mitophagy during the mitochondrial remodeling that occurs in WAT of obese mice (39). Further, our studies suggest that the previously described roles of parkin in prevention of DIO and maintenance of beige adipocytes (6,20,21) occur in an adipocyte-independent manner. These findings, in addition to those in ␤ cells above, place previous findings in whole-body knockouts in an appropriate cellular context and suggest a need to refine interpretations of genetic links between parkin and T2D in humans. In general, it is still not understood whether mitophagy is beneficial or detrimental in the progression of diseases such as cancer or metabolic syndrome (34). Our data to date suggest that parkin deficiency is dispensable for adipocytes and pancreatic ␤ cells in the regulation of wholebody metabolism.
Our findings agree with recent publications describing mild phenotypes when parkin or PINK1 is depleted in vivo (34 -36). The emergence of these studies places the importance of parkin in physiologically relevant contexts into question and suggests a potential for parkin-independent mitophagy pathways to compensate for maintenance of mitochondrial quality control. Mitophagy still occurs in mice lacking PINK1 or in Drosophila with either PINK1 or parkin deficiency, suggesting that other pathways maintain mitochondrial homeostasis despite their absence (34 -36). These parkin-independent pathways may include receptor-mediated mitophagy (including BNIP3, NIX/ BNIP3L, or FUNDC1 among others), lipid-mediated mitophagy (via cardiolipin on the inner mitochondrial membrane), E3 ubiquitin ligases (such as MUL1), or ubiquitin-binding protein (34). Further work is needed to better understand mechanisms by which these pathways compensate for the absence of parkin. For instance, how and under what conditions are specific mitophagy pathways activated to maintain healthy mitochondrial function (34)? Parkin may also have broader functions, as recent reports suggest roles in cellular processes unrelated to mitophagy (40). These pathways remain poorly understood but will likely be a major focus of future studies.

Parkin function in ␤ cells and adipocytes
This study offers crucial contributions to the study of metabolic diseases by highlighting that parkin, a T2D-associated gene and crucial regulator of mitophagy, is not necessary during overnutrition to control metabolic phenotypes in pancreatic ␤ cells or adipose tissue. Loss of parkin subtly alters lipid uptake and ␤-oxidation in cultured adipocytes and mildly impairs mitochondrial turnover in ␤ cell lines; however, this is not sufficient to disrupt whole-body glucose metabolism. Further study will be essential to dissect alternative regulators of mitophagy in pancreatic ␤ cells and adipocytes and their importance for development of T2D.

Animals
Prkn FL/FL (Parkin flox ) mice were a generous gift from Ted Dawson (Johns Hopkins University) and Lexicon Genetics and were generated with loxP sites flanking exon 7 of the Prkn allele (41). Pdx1-Cre mice were a generous gift from Doris Stoffers (University of Pennsylvania) (27). Ins1-Cre mice (026801) and Adiponectin-Cre mice (028020) were obtained from The Jackson Laboratory (Ellsworth, ME) (30,37). For DIO studies, mice were fed an HFD containing 60% calories from fat (Research Diets, 12492, New Brunswick, NJ). For beigeing studies, male mice were administered 1 mg/kg CL-316,243 intraperitoneally (Cayman Chemical, Ann Arbor, MI) once daily for 7 days, followed by a 15-day rest period without drug treatment. Female mice were placed in thermal chambers at 6°C (with a normal 12-h light cycle and free access to chow and water) for 3 weeks to induce beigeing, followed by 15 days at room temperature. All animal studies were performed in compliance with policies of the University of Michigan Institutional Animal Care and Use Committee.

Glucose tolerance tests and in vivo glucose-stimulated insulin secretion
For adipose tissue studies, animals were fasted for 16 h and then administered 1 mg/kg glucose intraperitoneally (IPGTT). For pancreatic tissue studies, animals were fasted for 6 h and then administered 2 mg/kg glucose intraperitoneally. Blood glucose concentrations were monitored 0, 15, 30, 60, and 120 min post-injection using Contour Next blood glucose strips (Bayer AG, Leverkusen, Germany). For glucose-stimulated insulin secretion, animals were fasted for 6 h, and then 3 mg/kg glucose was administered intraperitoneally. Glucose concentrations were measured, and plasma samples were collected 0 and 3 min post-injection. Plasma insulin concentrations were measured by ELISA (Alpco, Salem, NH).

Animal phenotyping
Body composition was measured by NMR spectroscopy using the LF90 II Minispec (Bruker, Billerica, MA). Food intake, activity, energy expenditure, and oxygen consumption were monitored for 3 days using the Comprehensive Lab Animal Monitoring System (Columbus Instruments, Columbus, OH).

Parkin function in ␤ cells and adipocytes
All animal phenotyping was performed by the University of Michigan Mouse Metabolic Phenotyping Core.
MIN6 ␤ cells were maintained as described previously (15). siRNA studies were carried out as described previously (43). Briefly, MIN6 ␤ cells were seeded on 6-well plates. 24 h later, they were treated with 2 M NT or Parkin-specific siRNA (Dharmacon, Lafayette, CO) using Dharmafect 3 transfection reagent (Dharmacon). Cells were cultured for 48 h before protein isolation or glucose-stimulated insulin secretion (GSIS) assays, which were performed as described previously (43).

Measurement of cellular ROS
Total cellular ROS was measured in MIN6 ␤ cells with or without parkin-specific siRNA or in Parkin flox -Ad-GFP and Parkin flox -Ad-Cre adipocytes using the a cellular ROS assay kit according to the manufacturer's instructions (Abcam, Cambridge, UK). For MIN6 assays, cells were seeded on black, clearbottom 96-well plates (Grenier Bio-One, Kremsmünster, Austria), and cellular ROS were assessed using a red fluorescence kit (Abcam, ab186027). For adipocyte assays, cells were seeded on 24-well black, clear-bottom plates (PerkinElmer Life Sciences, Turcu, Finland), and cellular ROS were assessed using only the ROS portion of the ROS/superoxide detection kit (Abcam, ab139476).

Adipocyte and stromal vascular cell fractionation
Using a protocol modified from Rodbell (44), white adipose tissue (inguinal and gonadal combined) was isolated from Parkin flox and AD-Parkin KO mice, minced with scissors, and digested with 1 mg/ml collagenase (type I; Worthington Biochemical, Lakewood, NJ) in Krebs-Ringer-HEPES buffer and 3% fatty acid-free BSA (Gold Biotechnology, St. Louis, NJ). After 1 h of digestion at 37°C, the cell suspension was filtered through 100-m cell strainers. Adipocytes and the stromal vascular fraction were separated by differential centrifugation (100 ϫ g for 8 min) and washed with Krebs-Ringer-HEPES buffer containing 3% BSA.

Histology
Tissues were fixed in 10% neutral buffered formalin for 24 h and processed for paraffin embedding by the University of Michigan Microscopy and Imaging Analysis Core. Sections (5 m) were stained with hematoxylin and eosin as described previously (45). Pancreata were harvested and fixed in 4% paraformaldehyde overnight and either processed for paraffin embedding as above or incubated in 50% sucrose overnight and processed in optimal cutting temperature compound (OCT; Thermo Fisher Scientific) for cryosections. Immunostaining for insulin (Dako (Agilent), Santa Clara, CA) and glucagon (Santa Cruz Biotechnology Inc., Dallas, TX) was performed as described previously (15).

Immunoblot analysis
Immunoblots were performed as described previously (15,46). In brief, 5-20 g of cell or tissue protein extract was separated by SDS-PAGE, transferred onto PVDF or nitrocellulose membranes, and immunoblotted with primary antibodies (listed in Table S1).

Quantitative RT-PCR
Total RNA was isolated from frozen tissue or isolated cells using RNA STAT-60 (AMS Biotechnology, Cambridge, MA) according to the manufacturer's instructions. Reverse transcription and qRT-PCR were performed as described previously (46). To assess mitochondrial number, total RNA was treated with DNase and reverse-transcribed, and the expression of mitochondrial genes relative to nuclear genes was measured by qRT-PCR. A list of qRT-PCR primers is listed in Table S2.

Transmission EM
Brown adipose tissue from Parkin flox and AD-Parkin KO mice was minced into small fragments and fixed in 2.5% glutaraldehyde in Sorensen's phosphate buffer (pH 7.4) overnight at 4°C. Samples were washed in Sorensen's buffer, post-fixed in 2% osmium tetroxide in Sorensen's buffer for 1 h at room temperature, and then washed again with Sorensen's buffer and dehydrated through ascending concentrations of acetone before embedding in epoxy resin. Semithin sections (500 nm) were stained with toluidine blue for tissue identification. Selected regions of interest were sectioned at 70 nm in thickness and post-stained with uranyl acetate and Reynolds lead citrate. The sections were examined using a JEOL JEM-1400 Plus transmission electron microscope at 80 kV with support from the University of Michigan Microscopy and Imaging Analysis Core.