Systemic deficiency of PIKfyve, the evolutionarily conserved phosphoinositide kinase synthesizing cellular PtdIns5P and PtdIns(3,5)P2 and implicated in insulin signaling, causes early embryonic death in mice. In contrast, mice with muscle‐specific Pikfyve disruption have normal lifespan but exhibit early‐age whole‐body glucose intolerance and muscle insulin resistance, thus establishing the key role of muscle PIKfyve in glucose homeostasis. Fat and muscle tissues control postprandial glucose clearance through different mechanisms, raising questions as to whether adipose Pikfyve disruption will also trigger whole‐body metabolic abnormalities, and if so, what the mechanism might be. To clarify these issues, here we have characterized two new mouse models with adipose tissue disruption of Pikfyve through Cre recombinase expression driven by adipose‐specific aP2‐ or adiponectin (Aq) promoters. Whereas both mouse lines were ostensibly normal until adulthood, their glucose homeostasis and systemic insulin sensitivity were severely dysregulated. These abnormalities stemmed in part from accelerated fat‐cell lipolysis and elevated serum FFA. Intriguingly, aP2‐Cre‐PIKfyvefl/fl but not Aq‐Cre‐PIKfyvefl/fl females had severely impaired pregnancy‐induced mammary gland differentiation and lactogenesis, consistent with aP2‐Cre‐mediated Pikfyve excision in nonadipogenic tissues underlying this defect. Intriguingly, whereas mammary glands from postpartum control and Aq‐Cre‐PIKfyvefl/fl mice or ex vivo mammary gland explants showed profound upregulation of PIKfyve protein levels subsequent to prolactin receptor activation, such increases were not apparent in aP2‐Cre‐PIKfyvefl/fl females. Collectively, our data identify for the first time that adipose tissue Pikfyve plays a key role in the mechanisms regulating glucose homeostasis and that the PIKfyve pathway is critical in mammary epithelial differentiation during pregnancy and lactogenesis downstream of prolactin receptor signaling.
Mounting evidence implicates the enzymes catalyzing synthesis and turnover of phosphoinositides (PI) in human diseases, such as diabetes, cancer, neurological disorders, and others (Di Paolo and De Camilli 2006; Yuan and Cantley 2008; Hakim et al. 2012; Mayinger 2012; Shisheva 2012; Balla 2013; Sun et al. 2013). Although most mammalian PI‐metabolizing enzymes have a number of isoforms with a redundant function, the evolutionarily conserved PIKfyve (PhosphoInositide Kinase for position five containing a fyve domain) (gene symbol PIKFYVE), responsible for synthesis of nearly the entire cellular amounts for two of the seven PIs, that is, PtdIns5P and PtdIns(3,5)P2, is a product of a single gene (Shisheva 2008b). The low‐abundance PtdIns5P and PtdIns(3,5)P2 are signaling lipids, regulating various aspects of essential cellular processes such as endomembrane homeostasis, endosome processing in the course of membrane trafficking, insulin‐regulated cytoskeletal dynamics, autophagy, nuclear transport, gene transcription, and cell cycle (Shisheva 2012; Viaud et al. 2014; Shisheva et al. 2015; Vicinanza et al. 2015). The nonredundant function of PIKfyve is evident by the early embryonic lethality of Pikfyve null mice, occurring at the preimplantation (Ikonomov et al. 2011) or postimplantation stage (Takasuga et al. 2013). The phenotype severity of the systemic Pikfyve disruption is proportional to the level of the remaining protein/activity. For example, genetically modified mice expressing ~10% residual PIKfyve survive up to 2 weeks after birth, having various defects in several tissues, whereas heterozygous PIKfyve−/+ mice that exhibit ~50% residual protein/activity in all tissues, including brain, have a normal lifespan without ostensible tissue/organ abnormalities (Shisheva et al. 2015). In humans, the PIKFYVE mutations identified thus far are all heterozygous and outside the catalytic domain, triggering only mild corneal dystrophy (Li et al. 2005). Thus, combined data in humans and genetically modified mice suggest that key mutations in the PIKFYVE gene might be incompatible with human life, leading, presumably, to a demise of the developing embryo. Furthermore, consistent with the requirement for a tight control in PtdIns(3,5)P2/PtdIns5P production, PIKfyve forms a multiprotein complex incorporating the phosphatidylinositol 5‐phosphatase Sac3 (Sac1 domain‐containing phosphoinositide 5‐phosphatase 3, gene symbol FIG4) that not only turns over PtdIns(3,5)P2/PtdIns5P but also activates PIKfyve (Sbrissa et al. 2007; Ikonomov et al. 2009a). Concordantly, mutations in Sac3 are linked to human disease (Lenk and Meisler 2014).
PIKfyve was first identified in the context of muscle and fat tissues, both of which harbor the insulin‐regulated glucose transporter GLUT4 (Shisheva et al. 1999). The role of PIKfyve and its lipid products as part of the signaling cascade triggering insulin‐stimulated glucose transport and GLUT4 cell‐surface accumulation was subsequently confirmed in cultured 3T3L1 adipocytes and primary fat cells (Ikonomov et al. 2007, 2009b; Koumanov et al. 2012). Mechanistically, PIKfyve functions downstream of the PI3Ks that produce, directly or indirectly, its enzymatic substrate, that is, PtdIns3P, therefore, processes controlled by PIKfyve signaling would require normal PI3K (phosphoinositide 3‐kinase) functioning (Shisheva 2012; Ikonomov et al. 2015). Recent data in genetically modified mice (MPIfKO) with PIKfyve deficiency localized specifically in skeletal muscle, the tissue responsible for the majority of insulin‐regulated postprandial glucose disposal (Biddinger and Kahn 2006), has provided the first in vivo evidence for the central role of muscle PIKfyve in the mechanisms regulating whole‐body glucose homeostasis (Ikonomov et al. 2013). The MPIfKO mice exhibited systemic glucose intolerance and impaired insulin‐induced glucose uptake in muscle as well as increased adiposity and hyperinsulinemia but not dyslipidemia (Ikonomov et al. 2013). Thus, the combined phenotype manifested by the MPIfKO mouse recapitulates the cluster of typical features in clinical insulin resistance, also known as prediabetes, that includes systemic glucose intolerance and insulin resistance, hyperinsulinemia, and increased visceral obesity without dyslipidemia (Abdul‐Ghani and DeFronzo 2009).
The direct role of adipose tissue to the postprandial glucose clearance is minor (Biddinger and Kahn 2006; Herman and Kahn 2006). Nevertheless, it contributes to the development of prediabetes and T2D due to a failure of its two key functions: to store lipids as TG and to secrete bioactive factors (Yu and Ginsberg 2005; Jensen 2006; Deng and Scherer 2010). The current study was designed to explore for the first time the significance of adipose PIKfyve in glucose homeostasis. To this end, we initially crossed our PIKfyvefl/fl mice (Ikonomov et al. 2011) with a mouse line expressing Cre recombinase driven by the widely used 5.4 kb promoter/enhancer of the aP2/FABP4 (adipose protein 2/fatty acid binding protein 4) gene (He et al. 2003). Whereas we found that the PIKfyvefl/fl,aP2‐Cre+ mice of both genders had dysregulated glucose homeostasis, the females also exhibited severe defects in mammary gland differentiation during pregnancy and in lactogenesis, causing early postnatal death of the offspring in primiparity or multiparity. To further explore the role of adipose PIKfyve in glucose homeostasis as well as to reveal if PIKfyve‐deficient adipose tissue underlies the striking mammary phenotype, we generated another fat tissue‐specific mouse line, PIKfyvefl/fl,Aq‐Cre+, in which Cre‐recombinase is driven by the promoter of the most adipocyte‐specific gene, that is, adiponectin (Wang et al. 2010; Eguchi et al. 2011). Although PIKfyvefl/fl,Aq‐Cre+ mice exhibited severely impaired glucose homeostasis, the mammary gland phenotype was not reproduced, consistent with aP2‐Cre‐mediated Pikfyve excision in nonadipogenic tissues causing the impaired mammogenesis and defective lactogenesis. Furthermore, in search of the mechanisms underlying impaired glucose homeostasis in PIKfyvefl/fl,Aq‐Cre+ mice, we established accelerated fat‐cell lipolysis and elevated serum FFA (free fatty acids), thus revealing an unexpected role of adipose Pikfyve in the mechanisms regulating fat tissue TG storage.
Animal numbers and all animal procedures were approved by the Wayne State University Animal Care and Use Committee (IACUC). The study was conducted according to institutional ethics guidelines for animal work by the Association and Accreditation of Laboratory Animal Care.
PIKfyvefl/fl,aP2‐Cre+ and PIKfyvefl/fl,Aq‐Cre+ mice
The development of conditional PIKfyvefl/fl mice on a C57BL/6J background by the Cre‐loxP approach was described elsewhere (Ikonomov et al. 2011). To generate mice with selective disruption of Pikfyve in adipose tissue, the PIKfyvefl/fl progeny was crossed with mice expressing Cre recombinase under the adipose tissue‐specific promoters of aP2 (generated by Ronald Evans and purchased from The Jackson Laboratory; B6.Cg‐Tg(Fabp4‐cre)1Rev/J) and of Adiponectin (a kind gift by Dr. Evan Rosen, supplied by Dr. Richard M Mortensen on a C57BL/6J background). Genotyping was performed by PCR using genomic DNA isolated from tails of 20‐day‐old mice and specific primer‐pairs detailed previously (Ikonomov et al. 2011). PIKfyvefl/fl,aP2‐Cre+ and PIKfyvefl/fl,Aq‐Cre+ mice were used as homozygous mutants; PIKfyvefl/fl mice of the respective breeding were used as controls and PIKfyvefl/wt,aP2‐Cre+ or PIKfyvefl/wt,Aq‐Cre+ were used as heterozygous mutants. Most of the experiments were performed by comparing PIKfyvefl/fl controls versus the respective PIKfyvefl/fl,aP2‐Cre+ and PIKfyvefl/fl,Aq‐Cre+ mice. Animals were maintained in a temperature‐controlled environment (22 ± 1°C) with a 12‐h light/dark cycle (6 am/6 pm) and a standard rodent diet (4.5% calories from fat; LabDiets #5001) with free access to food and water. Separate groups of male and female mice were used as indicated.
Body composition and blood/serum metabolites
Fat and lean mass was measured in randomly fed nonanesthetized mice, typically between noon and 2 pm, using EcoMRI (Echo Medical systems 130). Glucose was determined by a glucometer (Truetrack) in a blood droplet formed after mouse tail clipping. Nonesterified fatty acids (FFA) and glycerol concentrations were measured in serum of 4 h fasted animals using enzymatic kits from Wako (99475409, Richmond, VA) and Sigma (TR0100 kit, Saint Louis, MO). Adiponectin was evaluated in serum by WB. Serum was derived from blood collected by heart puncture after mouse euthanasia by cervical dislocation.
Glucose and insulin tolerance tests
For GTT (glucose tolerance test) and ITT (insulin tolerance test), individually housed (on paper bedding), and morning‐fasted mice (5 h) were injected intraperitoneally (i.p.) with 1.5 g dextrose/kg body weight or insulin (0.75 U/kg body weight) (Humalog, Lilly). Tail blood prior to, and following glucose or insulin injections was used for glucose measurements at the indicated time intervals. Excel (Microsoft Inc.) and the trapezoid rule were used to calculate the area under the curve (AUC) from time zero to the last time point (120 min in GTT and 60 min in ITT) postinjection as previously described (Ikonomov et al. 2013).
In vivo insulin stimulation
Following 14 h fasting, mice were injected i.p. with either saline or 5 U/kg body weight of insulin. Five‐min postinjections mice were sacrificed by cervical dislocation. Adipose tissue was removed and frozen in liquid N2 for immunoblot analyses of the insulin signaling proteins.
Fat cell isolation and lipolysis
Fat cells from freshly isolated perigonadal fat were prepared by collagenase (type I; 134 unit/mg) digestion in KRH (Krebs‐Ringer‐Hepes) buffer, pH 7.4, as we previously described (Shisheva and Shechter 1992, 1993). Briefly, duplicate samples of fat cell suspension (~3 × 105 cells/mL, in KRH buffer containing 0.7% BSA) were treated with isoproterenol (25 μmol/L) (Sigma) for 60 min at 37°C. When insulin (human recombinant, a gift by Eli Lilly) was present, it was added 5 min prior to the incubation with isoproterenol. At the end of the incubation, glycerol was measured enzymatically (TR0100 kit from Sigma, Saint Louis, MO) at 540 nm in duplicate aliquots taken from the aqueous phase of the reaction mix.
Antibodies and Western blotting analyses
Rabbit polyclonal anti‐PIKfyve (R7069; East‐Acres, MA) (Sbrissa et al. 1999), anti‐ArPIKfyve (Associated regulator of PIKfyve) (WS047; Covance, Denver, PA) (Sbrissa et al. 2004a), anti‐Sac3 (Sbrissa et al. 2007) and anti‐GDI1 (GDP dissociation inhibitor 1) antibodies (Shisheva et al. 1994) (R5057; East‐Acres, MA) were produced and characterized previously. Anti‐phosphoThr308‐Akt (#9275), anti‐phosphoSer473‐Akt (#9271), anti‐Akt, and anti‐phosphoTyr694‐Stat5 (#4322) antibodies were from Cell Signaling (Beverly, MA). Polyclonal antibodies recognizing both Stat5a and Stat5b (Signal transducer and activator of transcription 5) were from Santa Cruz Biotechnology (sc‐835). Anti‐adiponectin antibodies were a kind gift by Dr. Phil Scherer. Western blotting was conducted as described (Ikonomov et al. 2011). Briefly, harvested mouse tissue was homogenized in RIPA+ (radioimmuoprecipitation assay) buffer (50 mmol/L Tris/HCl buffer, pH 8.0, containing 150 mmol/L NaCl, 1% Nonidet P‐40, 0.5% Na deoxycholate), supplemented with 1 × protease inhibitor mixture (1 mmol/L phenylmethylsulphonylfluoride, 5 μg/mL leupeptin, 5 μg/mL aprotinin, 1 μg/mL pepstatin, and 1 mmol/L benzamidine) or RIPA2+ buffer, containing in addition 1 × phosphatase inhibitor mixtures (50 mmol/L NaF, 10 mmol/L Na pyrophosphate, 25 mmol/L Na β‐glycerophosphate and 2 mmol/L Na metavanadate). Lysate protein was resolved by SDS‐PAGE, transferred onto nitrocellulose membrane and probed with the indicated antibodies.
Immunoprecipitation and in vitro lipid kinase activity
Mouse tissue lysates, clarified by centrifugation, were subjected to immunoprecipitation with anti‐PIKfyve antibodies as previously described (Ikonomov et al. 2013). Immunoprecipitates were subjected to Western blotting or to in vitro PIKfyve lipid kinase assays followed by TLC (thin layer chromatography) lipid resolution as detailed elsewhere (Ikonomov et al. 2013).
Mammary glands (#4 inguinal) were dissected from female mice at postpartum day 3 and fixed in 4% formaldehyde overnight at room temperature. The samples were then refixed in 1:1:1 (v/v) fixative, consisting of aqueous 2% osmium tetroxide, 2.5% glutaraldehyde in 0.1 mol/L phosphate buffer, and 0.2 mol/L Sorenson's phosphate buffer, in an ice bath for 1.5 h, then dehydrated in increasing concentrations of ethanol, and finally infiltrated in Epon‐Araldite plastic as described previously (Li et al. 2014). Thin sections were cut and stained in modified Richardson's stain as specified earlier (Li et al. 2014).
Mammary gland explants
Cultures of mammary gland explants were prepared from midpregnant (12–14 days of pregnancy) Swiss‐Webster mice (Harlan Laboratories, Indianapolis, IN) as previously described (Rillema 1994). Briefly, inguinal and thoracic mammary glands from 6 to 12 mice were removed, cut into small pieces and placed on siliconized lens paper floating on 6 mL M199 medium, containing insulin (1 μg/mL) and cortisol (10−7 mol/L). Following 24 h incubation at 37°C in an atmosphere of 95% O2‐5% CO2, the explants were treated without or with prolactin (1 μg/mL) for an additional 24 h. RIPA2+ lysates were obtained and analyzed by Western blotting with or without immunoprecipitation.
Data from PIKfyvefl/fl,aP2‐Cre+ and PIKfyvefl/fl,Aq‐Cre+ mice were compared with those obtained for flox/flox littermates or age‐matched mice of the corresponding cross by Student's t test. Statistical significance of ITT and GTT curves was determined by comparing differences in AUC of the groups by Student's t test. The films from Western blots and TLC were scanned at 300 d.p.i. for quantitative analysis. Several films of different exposure times were scanned and quantified to assure that the signals were within the linear range. All data are presented as mean ± SEM. Statistical significance is considered when P < 0.05.
Validation of the PIKfyvefl/fl,aP2‐Cre+ mouse model
The PIKfyvefl/fl mouse line (Ikonomov et al. 2011) was mated with mice carrying Cre recombinase under the aP2 promoter. Tail‐DNA genotyping of the F1–F3 progeny using PCR primers for Cre or floxed alleles identified mice with PIKfyvefl/fl,aP2‐Cre+ and PIKfyvefl/wt,aP2‐Cre+ genotypes of both genders for subsequent experimentation. PIKfyvefl/fl,aP2‐Cre+ mice were born at the expected Mendelian ratio. They were apparently normal, healthy and fertile, with a life span similar to that of their littermate PIKfyvefl/fl brothers and sisters up to 10 months of age, when the study ended.
Western blot analyses with anti‐PIKfyve antibodies revealed that PIKfyve protein levels were decreased by ~50–60% in perigonadal fat pads of both male and female PIKfyvefl/fl,aP2‐Cre+ mice versus corresponding PIKfyvefl/fl littermates (Fig. 1A). Considering the confounding effect by other cell types present in fat tissue, the Pikfyve gene inactivation was at reasonable efficacy. Data obtained with immunopurified PIKfyve preparations derived from periovarian and epididymal fat also revealed that the PIKfyve enzymatic activity was similarly reduced in PIKfyvefl/fl,aP2‐Cre+ mice as evidenced by the selective diminution in in vitro generated PtdIns(3,5)P2 and PtdIns5P products (Fig. 1B). Fat tissue of PIKfyvefl/wt,aP2‐Cre+ heterozygotes also exhibited reduced synthesis of the two lipid products, though to a lesser extent compared to PIKfyvefl/fl,aP2‐Cre+ mice, as expected for haploinsufficiency (Fig. 1B). In contrast, fat‐tissue levels of the PIKfyve‐associated phosphoinositide 5‐phosphatase Sac3 (Sbrissa et al. 2007; Ikonomov et al. 2009a) remained unaltered (Fig. 1A), validating that we monitor solely the consequences of Pikfyve inactivation. Notably, whereas in our conditional model, the aP2 promoter did not confer recombination in other metabolic tissue such as muscle and liver (Fig. 1C), there was a pronounced (40–50%) loss of PIKfyve, but not of Sac3, in the brain (Fig. 1D) in agreement with other observations using aP2‐Cre‐driven recombination (Urs et al. 2006; Martens et al. 2010; Heffner et al. 2012; Zhang et al. 2012). This reduction was apparent in both male and female PIKfyvefl/fl,aP2‐Cre+ mice and was evident even in brains derived from PIKfyvefl/wt,aP2‐Cre+ heterozygotes (Fig. 1D). It should be emphasized, however, that despite the PIKfyve reduction in brain, the PIKfyvefl/fl,aP2‐Cre+ mice of both sexes were normal in terms of body weight (Fig. 2A and E), fertility (Table 1), organ morphology, and organ size (not shown). This observation is reminiscent of the heterozygous PIKfyveWT/KO mice with Pikfyve gene inactivation at the whole‐body level, which, while exhibiting ~50% reduction in PIKfyve protein or activity in the brain, were ostensibly normal (Ikonomov et al. 2011).
PIKfyvefl/fl,aP2‐Cre+ mice are glucose intolerant and insulin resistant
Although female PIKfyvefl/fl,aP2‐Cre+ mice were slightly lighter at weaning, their body weight rapidly advanced and after 6 weeks, they reached the weight of their PIKfyvefl/fl littermates (Fig. 2E). The growth rate for body weight of male PIKfyvefl/fl,aP2‐Cre+ and PIKfyvefl/fl littermate controls were similar (Fig. 2A). To determine the systemic metabolic consequences of PIKfyve‐deficient adipose tissue, we performed glucose‐ and insulin‐tolerance tests (GTT and ITT) in 3‐month‐old male and female PIKfyvefl/fl,aP2‐Cre+ mice and compared them with PIKfyvefl/fl sex‐ and age‐matched controls. GTT in male mice revealed that, whereas blood glucose levels at 0 min were similar between the control and mutant groups, they were greater in the PIKfyvefl/fl,aP2‐Cre+ mice subsequent to glucose injection (Fig. 2B). Similarly, higher glucose levels during GTT were seen with the female PIKfyvefl/fl,aP2‐Cre+ compared to PIKfyvefl/fl littermate mice (Fig. 2F). In both genders, the difference became statistically significant at later time points, consistent with the idea for defective glucose clearance by the muscle (or fat) rather than suppression of endogenous glucose production by the liver (or kidney). Concordantly, the glucose area under the curve (AUC) during GTT for both males and females was larger in the PIKfyvefl/fl,aP2‐Cre+ genotype compared to age‐matched controls, reaching statistical significance for the female mice (Fig. 2D and H).
Insulin sensitivity in 3‐month‐old male or female PIKfyvefl/fl,aP2‐Cre+ mice was also impaired as judged by the blunted hypoglycemic response to i.p. injection of insulin during ITT. Thus, as illustrated in Fig. 2C and G, blood glucose levels in PIKfyvefl/fl,aP2‐Cre+ mice remained higher compared to those in sex‐ and age‐matched control mice at any single time point. Of note, basal glucose levels between the two groups of each sex were not statistically different (not shown). The reduced insulin sensitivity was further underscored by the AUC during ITTs, which was significantly larger in both male and female PIKfyvefl/fl,aP2‐Cre+ mice compared to the respective age‐ and sex‐matched control PIKfyvefl/fl mice (Fig. 2D and H). Together, these data demonstrate that both male and female PIKfyvefl/fl,aP2‐Cre+ mice exhibit systemic glucose intolerance and insulin resistance.
Validation of the PIKfyvefl/fl,Aq‐Cre+ mouse model
The substantial reduction in PIKfyve levels in brain of both male and female PIKfyvefl/fl,aP2‐Cre+ mice suggests the possibility that dysregulated glucose homeostasis could be adversely associated with decreased PIKfyve in brain rather than in adipose tissue. Therefore, to identify the relative contribution of adipocyte PIKfyve to the observed dysregulation of glucose homeostasis as well as to begin exploring the potential mechanisms for the systemic glucose intolerance and insulin resistance, we have generated another fat‐tissue specific Pikfyve KO (knockout) model, in which expression of Cre recombinase is driven by the promoter of adiponectin, a gene that is strictly adipose‐specific (Wang et al. 2010; Eguchi et al. 2011; Lee et al. 2013). As PIKfyvefl/fl,aP2‐Cre+, the PIKfyvefl/fl,Aq‐Cre+ mice were born at the expected Mendelian ratio. They were ostensibly normal, healthy, and fertile (Table 1), with growth curves for body weight (Fig. 3A and D), being similar to those of their littermate PIKfyvefl/fl brothers and sisters up to at least 10 months of age.
There were marked decreases in PIKfyve protein levels as well as in PIKfyve lipid kinase activity (by ~80%) in perigonadal fat pads isolated from both male and female PIKfyvefl/fl,Aq‐Cre+ mice versus the corresponding control PIKfyvefl/fl littermates as revealed by Western blotting and in vitro lipid kinase assays (Fig. 3B, C, and E). Levels of the Sac3 phosphatase remained unaltered (Fig. 3B). In other fat depots of both genders, such as BAT or subcutaneous fat, PIKfyve was also profoundly reduced, as illustrated herein by dramatically decreased lipid kinase activity of immunopurified PIKfyve preparations (Fig. 3C and E). Notably, we did not observe any reduction in PIKfyve protein levels in brain of PIKfyvefl/fl,Aq‐Cre+ mice (Fig. 3B), nor were there changes in skeletal muscle in both males and females, as measured by the in vitro lipid kinase activity in anti‐PIKfyve immunoprecipitates (Fig. 3C and E).
PIKfyvefl/fl,Aq‐Cre+ mice are glucose intolerant and insulin resistant
Both male and female PIKfyvefl/fl,Aq‐Cre+ mice exhibited profoundly dysregulated glucose homeostasis and systemic insulin resistance as evidenced by the GTT and ITT in 6‐month‐old mice. Thus, as illustrated in Figure 4A, B, D, and E, glucose levels remained markedly higher subsequent to i.p injection of either glucose or insulin, with statistically significant differences at each time point. In contrast, basal levels of fasting glucose did not statistically differ between the two groups of each sex (Fig. 4A, B, D, and E). Of note, the differences in glucose levels were statistically significant at the early as well as the late time points of GTT in both genders, consistent with the idea that both defective glucose clearance by the muscle (or fat) and suppression of endogenous glucose production by the liver (or kidney) may account for the defects. The profound systemic glucose intolerance and insulin resistance in PIKfyvefl/fl,Aq‐Cre+ mice are underscored upon calculating AUC during ITTs and GTTs (Fig. 4C and F). In both male and female PIKfyvefl/fl,Aq‐Cre+ mice these values were markedly greater compared to PIKfyvefl/fl,Aq‐Cre− mice (by ~30%) and these differences were with high statistical significance (Fig. 4C and F). The reproducibility of dysregulated systemic glucose tolerance and insulin resistance in the two mouse models is consistent with the notion that the PIKfyve reduction in the brain as observed in the PIKfyvefl/fl,aP2‐Cre+ mice may not contribute to the metabolic abnormalities manifested by the PIKfyvefl/fl,aP2‐Cre+ mice.
Normal adiponectin but increased serum FFA in PIKfyvefl/fl,Aq‐Cre+ mice
PIKfyve controls proper performance of a number of membrane trafficking pathways, including exocytosis or secretion of a plethora of ion channels, metabolite transporters, and other active molecules (Shisheva 2012). Therefore, we hypothesized that one mechanism by which PIKfyve deficiency in fat may trigger systemic glucose intolerance and insulin resistance is associated with reduced plasma levels of adiponectin, a hormone predominantly secreted by the fat tissue (Halberg et al. 2008). Reduced plasma concentration of adiponectin, and particularly that of high molecular weight (HMW) forms, has been previously shown to accelerate insulin resistance (Pajvani et al. 2004). However, as illustrated in Figure 5A, our examination of the different forms of serum adiponectin by SDS‐PAGE under reducing or nonreducing conditions and immunoblotting with adiponectin antibodies failed to reveal changes in PIKfyvefl/fl,Aq‐Cre+ versus PIKfyvefl/fl,Aq‐Cre− mice. Next, as increased circulating levels of FFA could powerfully dysregulate systemic glucose homeostasis by decreasing insulin sensitivity of other metabolic organs (Guilherme et al. 2008; Kusminski et al. 2009), we assessed the fasting levels of circulating FFA and glycerol in 9‐month‐old PIKfyvefl/fl,Aq‐Cre+ and PIKfyvefl/fl,Aq‐Cre− mice. Intriguingly, we found that serum FFA were markedly increased in the PIKfyvefl/fl,Aq‐Cre+ versus PIKfyvefl/fl,Aq‐Cre− mice (Fig. 5B). These data indicate that the metabolic defects in PIKfyvefl/fl,Aq‐Cre+ could result in part from increased levels of circulating FFA.
Increased lipolysis and blunted antilipolysis by insulin in fat cells from PIKfyvefl/fl,Aq‐Cre+ mice
Elevated circulating FFA levels are often associated with increased fat mass. It is thus possible that whereas PIKfyvefl/fl,Aq‐Cre+ mice had body weight similar to littermate controls (Fig. 2A and E) they could have exhibited a relative expansion of adipose mass. Our analysis of total fat and lean mass by noninvasive EchoMRI in young‐adult and mature‐adult mice did not detect changes in adiposity of PIKfyvefl/fl,Aq‐Cre+ mice (Fig. 5C). Intriguingly, the perigonadal fat dissected from males or females at 9 months of age showed a trend for an increase by 5–12% in the mutant mice of both genders, but this difference did not reach statistical significance. Thus, whereas the trend for fat mass increases might, in part, underlie the elevated circulating FFA, it probably does not fully explain the defect.
Another cellular mechanism that could account for the observed greater levels of circulating FFA in PIKfyvefl/fl,Aq‐Cre+ mice is increased lipolytic activity of fat cells (McGarry 2002). Consistently, we found that fat cells freshly isolated from perigonadal fat of PIKfyvefl/fl,Aq‐Cre+ mice displayed markedly elevated TG hydrolysis versus PIKfyvefl/fl control mice upon β‐adrenergic receptor stimulation (isoproterenol) but the basal lipolysis was unaltered (Fig. 5D). Because insulin prominently inhibits lipolysis, we assessed its ability to counteract the lipolysis. As illustrated in Figure 5D, fat cells from of PIKfyvefl/fl,Aq‐Cre+ mice displayed blunted ability of insulin to restrain isoproterenol‐induced lipolysis compared to control PIKfyvefl/fl,Aq‐Cre− mice.
The dampened antilipolytic activity of insulin in the PIKfyvefl/fl,Aq‐Cre+ fat cells suggests that the latter might be insulin resistant. A major molecular event of the insulin signaling cascade is Akt activation, which mediates many insulin cellular responses, including antilipolysis and glucose entry (Huang and Czech 2007; Choi et al. 2010). Studies in cultured 3T3L1 adipocytes have revealed that reduced PIKfyve protein or enzymatic activity dampens insulin‐induced Akt phosphorylation (Ikonomov et al. 2002, 2007; Shisheva 2008a), with the PIKfyve product PtdIns5P found to be a potent activator (Pendaries et al. 2006; Ramel et al. 2009; Jones et al. 2013). Therefore, to reveal whether adipose tissue PIKfyve/PtdIns5P ablation affects the in vivo Akt phosphorylation in fat, we performed immunoblotting with phosphoAkt antibodies of adipose tissue lysates derived from insulin‐ or saline‐injected PIKfyvefl/fl,Aq‐Cre+ and PIKfyvefl/fl mice. As illustrated in Figure 5E, whereas 5 min post injection the AktS473 and AktT308 sites were profoundly phosphorylated in fat of control mice, their phosphorylation was greatly suppressed in fat from PIKfyvefl/fl,Aq‐Cre+ mice. These data indicate blunted Akt phosphorylation in fat tissue upon disruption of adipose Pikfyve, consistent with the requirement of adipose PIKfyve and its lipid products for efficient Akt phosphorylation by insulin.
Severe mammary gland phenotype by aP2‐Cre‐ but not by Aq‐Cre‐driven PIKfyve deficiency
While breeding the aP2‐Cre mouse line, we made the surprising observation that whereas completely fertile, giving birth to a similar number of pups as the heterozygous PIKfyvefl/wt,aP2‐Cre+ or control PIKfyvefl/fl littermates, the PIKfyvefl/fl,aP2‐Cre+ mothers failed to nurse their pups either from first (Table 1) or second and third pregnancies (in each, n = 3 PIKfyvefl/fl,aP2‐Cre+). However, the offspring from PIKfyvefl/fl females and PIKfyvefl/fl,aP2‐Cre+ males developed without problems. The pups born to female PIKfyvefl/fl,aP2‐Cre+ dams died 6 – 96 h postpartum. There was no milk in pups’ stomachs, visible through their translucent bodies, suggesting a failure of PIKfyvefl/fl,aP2‐Cre+ dams to lactate. Maternal instincts seemed preserved as judged by the fact that PIKfyvefl/fl,aP2‐Cre+ dams built a nest days before parturition and initially lay on top of the litter, but later, when anxiety took over, they abandoned the nest even when the pups were still alive (Movie S1). Because pups born to PIKfyvefl/wt,aP2‐Cre+ dams were nurtured normally (Table 1), an adverse effect of the Cre transgene was ruled out (Palmer et al. 2006; Robinson and Hennighausen 2011). Likewise, PIKfyvefl/fl,Aq‐Cre+ dams, in which adipose PIKfyve was even more efficiently eliminated than in PIKfyvefl/fl,aP2‐Cre+ (Figs. 1, 3) successfully nursed their litters (Table 1), thus, dismissing the possibility for PIKfyve deficiency in adipose tissue underlying the mammary gland defect during pregnancy.
The ability to nurse correlates with the extent of alveolar development and differentiation. Remarkably, at postpartum day 3 (PP3) both thoracic and inguinal glands in PIKfyvefl/fl,aP2‐Cre+ dams weighed markedly less (by 2.4 ± 0.5 fold and 6.9 ± 0.9 fold, respectively, and Figure S1) compared to those in PIKfyvefl/fl,aP2‐Cre− littermates, suggesting defective alveolar growth and differentiation during pregnancy. Strikingly, histological examination of mammary glands in PP3 PIKfyvefl/fl,aP2‐Cre+ mothers revealed a severely undeveloped lobulo‐alveolar system occurring at only ~15% density of the mammary fat pad and predominance of mammary adipocytes (Fig. 6A). The sparse alveoli‐like structures in the PIKfyvefl/fl,aP2‐Cre+ glands were also morphologically different in that they did not exhibit open lumen and were practically milk‐empty. In contrast, postpartum glands of control PIKfyvefl/fl,aP2‐Cre− females exhibited fully developed and distended milk‐filled alveoli, occupying the entire mammary fat pad (Fig. 6A). Remarkably, western blot analysis of mammary gland lysates derived from postpartum PIKfyvefl/fl,aP2‐Cre+ mice revealed practically undetectable levels of PIKfyve, a result further confirmed through concentrating the PIKfyve protein on anti‐PIKfyve antibodies and analyzing by immunoprecipitation (Fig. 6B). However, there were exceptionally high levels of PIKfyve levels in PP3 PIKfyvefl/fl,aP2‐Cre− mammary glands (Fig. 6B) not seen thus far in any other tissue. Quantification of a dilution series of mammary gland lysates by Western blotting revealed >20‐fold increase of PIKfyve levels in PP3 PIKfyvefl/fl,aP2‐Cre− versus PP3 PIKfyvefl/fl,aP2‐Cre+ dams. Notably, similar histology and Western blot analyses in mammary glands showed no substantial differences between PP3 PIKfyvefl/fl,Aq‐Cre+ and PIKfyvefl/fl,Aq‐Cre− mothers with respect to morphological appearance and density of the lobulo‐alveolar system within the mammary stroma (Fig. 6C) as well as PIKfyve protein levels (Fig. 6D).
Stat5a, a transcription factor that is phosphorylated downstream of the prolactin receptor (PrlR) activation, is the principal mediator of mammary gland development (Liu et al. 1997; Cui et al. 2004). Whereas it is expressed in all tissue, both total and phosphorylated Stat5 levels profoundly increase in mammary glands during late pregnancy and lactation. To reveal if abnormal Stat5 levels and/or activation occur in PIKfyvefl/fl,aP2‐Cre+ dams upon parturition, we performed Western blotting of mammary gland lysates with antibodies for total and phospho‐Stat5a/b. As illustrated in Figure 6E, there was a profound suppression in the surge of phospho‐ and total Stat5 levels in both thoracic and inguinal glands from PP3 PIKfyvefl/fl,aP2‐Cre+ versus littermate PIKfyvefl/fl,aP2‐Cre− mothers. Conversely, perigonadal Stat5 remained similar in mutant and control dams (Fig. 6E).
The striking increase in PIKfyve protein levels seen in lactating mammary glands of PIKfyvefl/fl,aP2‐Cre− or PIKfyvefl/fl,Aq‐Cre+ dams suggests the possibility that PIKfyve expression is markedly upregulated under the stimulus of pregnancy. We addressed this prediction by Western blotting with and without PIKfyve immunoprecipitation of lysates derived from both thoracic and inguinal mammary glands in nulliparous virgin, pregnant or postpartum day 3 lactating control PIKfyvefl/fl littermate females. This analysis revealed a slight PIKfyve increase in late pregnancy (by 2.5‐fold), but massive elevation after parturition (>15‐fold), concomitant with increased Stat5a/b activation (Fig. 6F). The surge in PIKfyve levels was paralleled by elevation in the PIKfyve complex, evidenced herein by the increased Sac3 phosphatase and ArPIKfyve scaffolding protein recovered in PIKfyve immunoprecipitates (Fig. 6G). Taken together, these data indicate that the inability of PIKfyvefl/fl,aP2‐Cre+ to respond to the stimulus of pregnancy is associated with arrested epithelial cell proliferation and lack of concomitant surge in PIKfyve levels. Our data are consistent with the notion that PIKfyve function is required in mammary epithelial differentiation and lactogenesis either as part of the PrlR‐Jak2‐Stat5 signaling pathway or as a parallel pathway.
Prolactin upregulates PIKfyve in an ex vivo mammary gland model of lactation
To delineate if PIKfyve's role in mammary gland development during pregnancy and lactation is mechanistically linked with the PrlR signaling pathway, we used mammary gland explants derived from midpregnant female mice, an ex vivo system that we have previously developed and characterized with respect to prolactin‐induced milk protein production (Rillema and Schneider‐Kuznia 1980; Rillema 1994). Subsequent to 24 h culturing in media containing insulin and cortisol, mammary gland explants were treated with or without prolactin (Prl) for additional 24 h. As illustrated in Figure 7A and B, Prl profoundly increased PIKfyve protein levels as revealed by western blotting with anti‐PIKfyve antibodies, with or without prior immunoprecipitation. The upregulation of PIKfyve coincided with a rise in total and phospho‐Stat5a/b (Fig. 7A). These data indicate that the PIKfyve pathway is activated downstream of PrlR stimulation and suggest that PIKfyve is required to confer the response to Prl during lactation.
Adipose PIKfyve is essential for glucose homeostasis
Due to the ability to store lipids and function as an endocrine organ, the adipose tissue controls insulin sensitivity of other metabolic organs, such as liver, muscle, and pancreas, which in turn, maintain glucose, and lipid homeostasis (Herman and Kahn 2006; Halberg et al. 2008). Therefore, pathophysiological conditions affecting fat tissue, including insulin resistance, ultimately impair systemic glucose tolerance and insulin sensitivity, leading to obesity‐related metabolic diseases. Whereas cell studies with PIKfyve protein knockdown or inhibition of the lipid kinase activity have established reduced insulin responsiveness of glucose uptake in 3T3L1 adipocytes (Ikonomov et al. 2007, 2009b; Shisheva 2008a; Koumanov et al. 2012), whether adipose tissue PIKfyve deficiency will impact systemic glucose homeostasis remains unknown. In this study, we addressed this question by taking advantage of a recently generated mouse model, in which the Pikfyve gene can be deleted in a tissue‐specific manner by the Cre‐loxP approach (Ikonomov et al. 2011, 2013). We developed two new mouse lines with Cre‐mediated inactivation driven by the promoters of the aP2/FABP4 or adiponectin genes, both of which confer recombination in adipose tissue, though with a different efficacy and specificity (Lee et al. 2013; Jeffery et al. 2014). We report here for the first time that mice with adipose PIKfyve deficiency exhibit systemic glucose intolerance and insulin resistance as judged by the accelerated rise and blunted fall of blood glucose in response to a systemic load of glucose and insulin, respectively (Figs. 2, 4).
The usage of two independent promoters for adipose‐specific gene disruption supports our conclusion that the primary defect underlying the whole‐body glucose intolerance and insulin resistance in PIKfyvefl/fl,aP2‐Cre+ and PIKfyvefl/fl,Aq‐Cre+ mice is the PIKfyve deficiency specifically targeted in adipose tissue. Data interpretation based solely on the aP2‐Cre‐driven recombination of adipose Pikfyve could have posed a significant level of uncertainty, as we, like others (Urs et al. 2006; Martens et al. 2010; Heffner et al. 2012; Zhang et al. 2012), have noted substantial PIKfyve ablation in the brain (Fig. 1). Disrupted neuronal function on its own could impair whole‐body response to a systemic glucose load and affect overall physiological control of blood glucose (Parton et al. 2007; Osundiji and Evans 2011). Therefore, to avoid confounding interpretations due to neuronal PIKfyve loss, we focused on the mechanism(s) by which adipose PIKfyve triggers metabolic dysfunction in the PIKfyvefl/fl,Aq‐Cre+ model that, in addition to higher specificity, exhibited also greater efficacy of PIKfyve ablation in fat tissue than the PIKfyvefl/fl,aP2‐Cre+ mice (Figs. 1, 3). Because the plasma adiponectin concentration is inversely associated with insulin resistance or T2D and because PIKfyve is a powerful regulator of exocytosis (Shetty et al. 2009; Shisheva 2012), we envisioned reduced adiponectin secretion by adipose tissue to underlie the observed metabolic defects. However, whereas such changes were not apparent (Fig. 5A), we found an unanticipated rise in circulating levels of FFA, with corresponding increases in serum glycerol (Fig. 5B). High circulating FFA have long been known to impair insulin sensitivity of muscle or liver and blunt insulin secretion by pancreatic beta cells, resulting in hyperglycemia (Haber et al. 2003; Herman and Kahn 2006; Jensen 2006; Ye 2007; Rhodes et al. 2013). It will be important in future studies to reveal the relative contribution for each of these tissues to the systemic glucose intolerance observed in PIKfyvefl/fl,Aq‐Cre+ mice.
An important insight in our study is the delineation of the cellular mechanism by which the loss of adipose PIKfyve causes elevation of circulating FFA. Thus, our findings for markedly augmented lipolytic activity in PIKfyvefl/fl,Aq‐Cre+ fat cells upon β‐adrenergic receptor stimulation (Fig. 5D) corroborate the notion that adipose PIKfyve deficiency impairs fat tissue TG storage, resulting in elevation of serum FFA levels. A potential signaling mechanism for accelerated fasted fat cell lipolysis may involve mTORC1 (mammalian target of rapamycin complex 1), whose reduced activation promotes TG hydrolysis (Chakrabarti et al. 2010; Soliman et al. 2010). Reportedly the PIKfyve lipid product PtdIns(3,5)P2 activates mTORC1 (Bridges et al. 2012), an observation consistent with the notion that the higher TG hydrolysis observed in the PIKfyvefl/fl,Aq‐Cre+ fat cells could be due at least in part to reduced mTOR1 activation. Furthermore, in the face of severe fat‐cell insulin resistance in PIKfyvefl/fl,Aq‐Cre+ mice, as evidenced in our study by the profound reduction in Akt activation in response to intraperitoneal injection of insulin, we observed reduced ability of insulin to restrain β‐adrenergic receptor‐activated lipolysis (Fig. 5D and E). Sustained insulin activation of Akt is negatively regulated by the PP2A phosphatase (Ugi et al. 2004; Liao and Hung 2010; Toker and Marmiroli 2014). Intriguingly, the PIKfyve lipid product PtdIns5P is a powerful suppressor of the PP2A activity (Ramel et al. 2009), which may explain the lower levels of phospho‐Akt under adipose PIKfyve deficiency. Reduced phosphoAkt may further blunt the mTORC1 activation (Lamming and Sabatini 2013; Chakrabarti and Kandror 2015) to deepen accelerated lipolysis in PIKfyvefl/fl,Aq‐Cre+ fat cells. Furthermore, in the presence of adipocyte insulin resistance, effective re‐esterification of FFA is likely to be hampered, as the source for the required supply of glycerol 3‐phosphate is the insulin‐sensitive glycolysis and/or glyconeogenesis, thus, further elevating FFA (Herman and Kahn 2006; Kusminski et al. 2009). It should be noted, however, that the observed elevation of serum FFA and increased fat cell lipolytic rate in PIKfyvefl/fl,Aq‐Cre+ mice may not be the only mechanisms by which adipose PIKfyve deficiency impairs whole‐body glucose homeostasis. This prediction is corroborated by our secretome analysis (Shisheva and Chen, in preparation), revealing increased secretion by the PIKfyvefl/fl,Aq‐Cre+ fat cells of a number of proinflammatory factors such as plasminogen activator inhibitor 1, haptoglobin, and thrombospondin1, which contribute to insulin resistance (Deng and Scherer 2010; Jung and Choi 2014).
Unexpected significance of PIKfyve to mammogenesis and lactogenesis
Whereas the aP2 promoter may be less adipose‐specific than originally thought, its off‐target expression in various tissues and organs has helped in disclosing important and unexpected functions or mechanisms (Zhang et al. 2012; Xiang et al. 2013). Our study provides another paradigm for such an unanticipated mechanism. Thus, we revealed a critical role for PIKfyve in mammary gland development, which emerged as an unintended consequence of aP2‐Cre driven pikfyve recombination in off‐target tissues. Our observation for severely reduced lobuloalveolar growth in PIKfyvefl/fl,aP2‐Cre+ females and the unconditional failure of dams to produce milk and nurse (Fig. 6A and Table 1) establishes an unforeseen function of PIKfyve in mammary epithelial differentiation, proliferation, and lactogenesis. Because the PIKfyvefl/fl,Aq‐Cre+ females did not have such a defect (Table 1), ablation of adipose tissue Pikfyve as an underlying cause is highly improbable. Likewise, systemic heterozygous PIKfyveWT/KO mice, with one half of the wild type gene/protein in all tissues (Ikonomov et al. 2011), readily nurse their offspring. These data suggest that the defect in the PIKfyvefl/fl,aP2‐Cre+ females is due to a highly efficient off‐target Pikfyve recombination. Because recent data indicate high promiscuity of aP2‐Cre expression at both pre‐ and postnatal stages (Heffner et al. 2012), the off‐target sites of Pikfyve inactivation triggering impaired mammogenesis was not addressed in our study. However, we surmised the mammary epithelial precursor cells as an off‐target candidate. Their propagation and differentiation during pregnancy and lactation is likely to require PIKfyve, as suggested by our observations for excessive increases of the mammary epithelium PIKfyve after parturition (Fig. 6C, F). The embryonic lethality of the global pikfyve disruption occurring as early as the preimplantation stage of the differentiation/developmental program due, in part, to arrested DNA synthesis/cell division (Ikonomov et al. 2011) also supports this prediction. Future studies using mammary gland‐specific promoters for Pikfyve excision will enlighten this issue.
A failure of mammary gland development and lactation during pregnancy, as observed in our PIKfyvefl/fl,aP2‐Cre+ model, is well‐characterized in female mice lacking individual members of the PrlR signaling cascade, including PrlR, Jak2 (Janus kinase 2), Stat5, and Stat5 regulators (Wagner and Rui 2008). Because PIKfyve was diminished in the brain of PIKfyvefl/fl,aP2‐Cre+ females (Fig. 1B), one might link the mammary gland defect with presumable Prl deficiency. Whereas reasonable, such a possibility seems unlikely in light of data indicating major infertility problems in female mice upon Prl inactivation (Horseman et al. 1997), contrary to our PIKfyvefl/fl,aP2‐Cre+ females, which readily conceived and produced normal litter sizes (Table 1). These observations together with other data in our study suggest that the PIKfyve pathway and the PrlR‐Jak2‐Stat5 signaling are interconnected. For example, concomitant with Stat5 activation, PIKfyve is also greatly upregulated in mammary epithelium subsequent to PrlR activation by the stimulus of pregnancy or by prolactin treatment of mammary gland explants (Figs. 6F, 7). Concordantly, postpartum PIKfyvefl/fl,aP2‐Cre+ females exhibited a marked reduction in Stat5 expression/activation in mammary glands (Fig. 6E). It should be emphasized, however, that in addition to PIKfyve deficiency, a lactation failure could also contribute to the deregulated Stat5 in PIKfyvefl/fl,aP2‐Cre+ females, thus making the relative contribution of each defect an important objective for future studies. Notwithstanding, our data in the mammary gland explants (Fig. 7) clearly indicate that PIKfyve is among the genes whose expression is regulated by PrlR activation. It is likely that PIKfyve, similar to the PI3K‐Akt in PrlR‐Jak2‐Stat5 signaling, may function as an upstream regulator of Stat5, downstream effector of Stat5, or both (Chen et al. 2010, 2012; Schmidt et al. 2014), which remains to be clarified in future research. In any case, our data indicate for the first time that PIKfyve and its lipid products, in addition to the signal‐transduction cascade of insulin (Sbrissa et al. 2004b; Ikonomov et al. 2007; Shisheva 2008a; Sarkes and Rameh 2010; Bridges et al. 2012; Koumanov et al. 2012), are also implicated in that of Prl.
The data presented herein reveal that adipose tissue PIKfyve is a key regulator of systemic glucose homeostasis. The usage of the aP2 promoter for Pikfyve ablation unraveled an unanticipated requirement for PIKfyve in mammary gland development and differentiation during pregnancy, downstream of prolactin‐receptor signaling.
Conflict of Interest
Figure S1. Littermates of the indicated genotype at postpartum day 3. Larger inguinal mammary glands (arrowheads) in the PIKfyvefl/fl,aP2‐Cre− versus PIKfyvefl/fl,aP2‐Cre+ dams.
Movie S1. PIKfyvefl/fl,aP2‐Cre+ dams built a nest days before parturition and initially lay on top of the litter, but later, when anxiety took over, they abandoned the nest even when the pups were still alive.
We thank Dr. Linda Hazlet, Ronald Barrett, and Larry Tait for their help with mammary glands histology. We thank Dr. Fred Miller for the helpful discussion of the mammary gland phenotype and Dr. Philipp Scherer, for the anti‐adiponectin antibodies. The senior author expresses gratitude to the late Violeta Shisheva for her many years of support.
↵† Equal contribution.
This research was supported by Wayne State University and WSU School of Medicine Research Offices, National Institute of Health (DK58058, to AS) and American Diabetes Association (7‐13‐BS‐161, to AS).
- Manuscript Received: April 29, 2016.
- Manuscript Accepted: May 4, 2016.
- © 2016 The Authors. Physiological Reports published by Wiley Periodicals, Inc. on behalf of the American Physiological Society and The Physiological Society.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.