Haploinsufficiency of the Insulin Receptor in the Presence of a Splice-Site Mutation in Ppp2r2a Results in a Novel Digenic Mouse Model of Type 2 Diabetes

Type 2 diabetes is a complex disease where cellular resistance to insulin combined with a failure in β-cell compensation results in the development of the disease. Underlying this process are multiple genetic and environmental factors that interact to determine susceptibility risk. However, there are relatively few examples of patients with diabetes whose disease can be demonstrated to be due to the interaction of mutations in two or more genes. One of these is due to heterozygous mutations in two unlinked genes, peroxisome proliferator–activated receptor γ (PPARG) and protein phosphatase 1, regulatory (inhibitor) subunit 3A (PPP1R3A), expressed in adipocytes and skeletal muscle, respectively, resulting in severe insulin resistance and lipodystrophy (1). A second example is haploinsufficiency for the insulin receptor (IR) in combination with chimerin 2 (CHN2), a GTPase-activating protein, that results in insulin resistance and deficiency in intrauterine growth (2). In this latter example, the CHN2 mutation implicates a novel gene in insulin signaling and its regulation of metabolism and growth (2). Although there are other examples of doubly heterozygous individuals with diabetes, e.g., in the maturity-onset diabetes of the young HNF1A and HNF4A genes, it is unclear how these impact the severity of disease (3). In a mouse model, a digenic insulin resistance phenotype has been described whereby 40% of mice heterozygous for both IR and insulin receptor substrate 1 (IRS-1) null alleles develop overt diabetes at 4–6 months of age, demonstrating how two mild impairments in the same pathway can interact to cause diabetes (4). The insulin signaling pathway, including IR, IRS, phosphoinositide 3-kinase, and AKT and its effectors, and pathways via extracellular signal–related kinase regulate key metabolic processes including gluconeogenesis, glucose uptake, glycogen synthesis, lipogenesis, and protein synthesis and growth (5–7). These highly regulated multistep pathways may be perturbed with multiple small effect mutations that collectively result in significant disruption and consequent disease (5).

Here, we describe a digenic mouse model of type 2 diabetes where haploinsufficiency of IR and an N-ethyl-N-nitrosourea (ENU)–induced novel splice-site mutation in the protein phosphatase 2A (PP2A), regulatory subunit B, α gene (Ppp2r2a) gives rise to a diabetic phenotype as a result of aberrant AKT signaling. We demonstrate the synergistic effect of two mutations affecting insulin signaling that leads to impaired glucose homeostasis when combined, supporting the concept that genetic susceptibility to diabetes can be determined by the interaction of small effect alleles.

Mice were kept in accordance with U.K. Home Office welfare guidelines and project license restrictions; in addition, the study was approved by the local Animal Welfare and Ethical Review Body. IR C57BL/6J knockout mice (8) were obtained from The Jackson Laboratory.

Genomic DNA was extracted from mouse tail or ear biopsy tissue using a Qiagen DNeasy Tissue Kit, and 250 ng was assayed against the Illumina Mouse Medium Density Linkage Panel (Illumina).

F1 founder genomic DNA 4 μg was fragmented by nebulization. The DNA-Seq library was further prepared from the fragmented DNA following the commercial instruction of a sample preparation for sequencing genomic DNA (Illumina). Next-generation sequencing analysis was performed on Array Suite software (Omicsoft). After DNA-Seq alignment using Omicsoft Aligner to mouse mm10, mutations were identified using Omicsoft’s Summarize Mutation function. The mutation report was further annotated with Ensembl gene models and dbSNP database. A list of interesting ENU mutation candidates was obtained by focusing on newly discovered missense mutations and mutations that could affect splicing events in ENU regions.

PCR primers for amplifying ENU-induced candidate hits in Ppp2r2a and integral membrane protein 2B (Itm2b) genes from genomic DNA were based on the sequences in GenBank. Primer sequences for Ppp2r2a were 5′-CAGTCCCTGTCTGTCTGTAACATACTCAG-3′ and 5′-CCCTTCCCACCAGATCACTCTTTGTC-3′ and for Itm2b were 5′-GCAAATTATCATATCTCTTTTGTCCGGATGCAC-3′ and 5′-GAATGTATATTTGAAGCTGGGCATGGCTG-3′. PCR-amplified gDNA or cDNA fragments were subcloned into the pCR II Vector using a TA Cloning Kit (Life Technologies) and sequenced with M13F and M13R primers or directly sequenced with the PCR primers on an ABI 3730xl DNA Analyzer.

Mice were tested using the European Mouse Phenotyping Resource of Standardised Screens (EMPReSS) simplified intraperitoneal glucose tolerance test (http://empress.har.mrc.ac.uk). Plasma glucose was measured using an Analox Glucose Analyzer GM9. For insulin tolerance tests, mice were fasted for 4 h and a baseline blood sample was taken followed by an intraperitoneal injection of 2 IU/kg of insulin. Blood samples were then taken at 10, 20, 40, and 60 min, and blood glucose was determined using an AlphaTRAK glucometer. Plasma insulin was measured using a Mercodia Mouse Insulin ELISA Kit. Mice were weighed at 2-week intervals between 12 and 30 weeks of age and were placed in metabolic cages (Tecniplast) for 24-h periods to measure food and water intake and urine output.

Mice were fasted overnight, given a surgical anesthetic dose (isoflurane) and 5 IU of insulin or saline injected directly into the hepatic portal vein, and killed 90 s later, and liver, gonadal fat pads, and gastrocnemius muscles were excised and immediately frozen in liquid nitrogen.

Hepa1-6, 3T3-L1, and C2C12 cells were purchased from ATCC and were cultured in DMEM (Invitrogen) supplemented with either 10% FBS (Invitrogen) or 10% calf serum (3T3-L1), 100 units ⋅ mL⁻¹ penicillin, and 100 mg ⋅ mL⁻¹ streptomycin (Invitrogen). Adipogenic differentiation of 3T3-L1 cells was induced by incubating cells in serum-free media for 48 h prior to supplementing the tissue culture medium with 250 μmol/L IBMX, 0.1 μmol/L dexamethasone, and 0.5 μg/mL insulin for 4 days. After this period, tissue culture medium was supplemented with insulin only.

Four small interfering RNAs (siRNAs) specific for Ppp2r2a were purchased from Qiagen, of which two oligos, 5′-TCCACGGAGAATATTTGCCAA-3′ (siRNA1) and 5′-AAGCATCACGAGAGAACAATA-3′ (siRNA3), gave greater than 70% knockdown. Stealth RNAi Negative Control Lo GC was purchased from Invitrogen. siRNA was transfected into Hepa1-6 or C2C12 cells at 60–70% confluency at a final concentration of 30 nmol/L in 6-well plates using Lipofectamine RNAiMax (Invitrogen). 3T3-L1 transfections were performed on day 8 of differentiation. Thirty hours after transfection, cells were serum starved for 18 h prior to incubation in serum-free media supplemented with 500 nmol/L insulin or saline for 15 min. Cell lysates were collected for either RNA or protein extraction.

Thirty hours after transfection with siRNAs, Hepa1-6 cells were incubated overnight in glucose-free DMEM media (0.1% BSA, 1 mmol/L sodium pyruvate). Cells were washed three times in PBS and incubated in glucose production buffer for 6 h (glucose-free DMEM without phenol red, 20 mmol/L sodium lactate, 2 mmol/L sodium pyruvate, 2 mmol/L L-glutamate, 15 mmol/L HEPES, 0.1% BSA) with or without 500 nmol/L insulin. Supernatant was collected and assayed for glucose concentration (Analox Glucose Analyzer GM9), cells were lysed, and protein concentration was quantified using a DC Protein Assay (Bio-Rad). Secreted glucose concentration was normalized to total protein concentration per well.

Protein was extracted from frozen tissue by homogenization in CelLytic protein lysis buffer (Sigma) supplemented with protease and phosphatase inhibitor cocktails (Roche). Protein was quantified using a DC Protein Assay (Bio-Rad).

Lysates of 1.3 mg/mL protein (3.75 µL) were mixed with 1.25 µL of Simple Western sample dilution buffer (ProteinSimple) containing a reducing agent and fluorescent standards to a final concentration of 1 mg/mL and denatured at 95°C for 5 min before analysis using an automated capillary electrophoresis system PEGGY. Primary antibodies against PPP2R2A(5689), AKT(9272), glycogen synthase kinase-3β(27C10) [GSK-3β(27C10)], ribosomal protein S6 kinase polypeptide 1(49D7) [p70S6K(49D7)] (Cell Signaling Technology), and tubulin (12G10 Developmental Studies Hybridoma Bank) were used in this study. Briefly, proteins were separated on the PEGGY instrument through a size-resolving matrix in capillaries, immobilized to the inner capillary wall, and incubated with primary and secondary antibodies before detection using chemiluminescence. Signal and quantitation of immunodetected proteins were generated automatically at the end of the run.

Total AKT, GSK-3β, and p70S6K and phosphorylated p70S6K(Thr-389), GSK-3β(Ser-9), AKT(Ser-473), and AKT(Thr-308) were quantified from 20 μg of total protein from cell lines and mouse tissues on Meso Scale Discovery (MSD) MULTI-ARRAY Assays K15133D-1, K15177D-1, and K151DYD-1.

Total RNA from frozen mouse tissues and/or Hepa1-6, C2C12, and 3T3-L1 cells were extracted using an RNeasy Plus Mini Kit (Qiagen). Quantitative RT-PCR using the TaqMan system (ABI Prism 7700) was carried out using cDNA generated by SuperScript III enzyme (Invitrogen), and gene expression was normalized relative to the expression of glyceraldehyde-3-phosphate dehydrogenase (Gapdh). TaqMan probes (Supplementary Table 1) were purchased from Applied Biosystems.

After exsanguination, pancreatic tissue was dissected, fixed in neutral buffered formaldehyde (Surgipath Europe Ltd., Bretton, U.K.), and longitudinally mounted in wax. Sections were cut and stained with hematoxylin and eosin.

Mouse line IGT10 was identified with impaired glucose tolerance from an ENU phenotype–driven screen sensitized by haploinsufficiency of the IR (leading to insulin resistance but not diabetes) as previously described (9) (Supplementary Fig. 1A). The F1 (C57BL/6J [ENU harboring] × C3H/HeH) male founder was backcrossed to C3H/HeH mice, and a cohort segregating both the IR knockout allele and random ENU-induced mutations were phenotyped in a glucose tolerance test at 12 and 24 weeks of age. Approximately 40% of the IR heterozygotes showed elevated fasted plasma glucose, elevated insulin, and glycosuria (Supplementary Fig. 1B–D).

The causative ENU mutation was mapped using single nucleotide polymorphism genotyping to a 7.7 Mb region of chromosome 14 between rs13482231 (66.978401) and rs6156908 (74.709292) (Fig. 1A). Next-generation sequencing of this region in the DNA from the F1 founder male identified only two ENU-induced mutations, confirmed by Sanger sequencing, both of which were noncoding. One was in intron 1 of the Itm2b gene with no predicted function (73.783471 A to G) and the second was in intron 3 of the Ppp2r2a gene (67.656803 T to G) predicted to create a new acceptor splice site (Fig. 1B and C). Use of the new splice site was predicted to result in the addition of eight amino acids followed by a premature stop (Supplementary Fig. 2). To test whether the new splice site was used, RNA was extracted from doubly IR/PPP2R2A and IR heterozygous mouse livers and quantitative RT-PCR was performed using a probe spanning exons 9-10. In doubly heterozygous IR/PPP2R2A mice, 55–65% of the Ppp2r2a transcript was correctly spliced compared with IR-only heterozygotes (Fig. 1D), and no difference was observed in Itm2b expression (Fig. 1E).

Additional phenotyping cohorts were generated by backcrossing to C3H/HeH. Fasted plasma glucose and insulin levels were measured every 2 weeks, and doubly heterozygous IR/PPP2R2A mice showed significantly elevated plasma glucose levels from 18 weeks compared with the other three groups (Fig. 2A) and the entire group exhibited glycosuria by 28 weeks of age (Fig. 2B). Plasma insulin levels were elevated in the IR and PPP2R2A heterozygous mice; in doubly heterozygous IR/PPP2R2A mice, there was a clear additive effect on increased insulin levels compared with wild-type littermates (Fig. 2C). Doubly heterozygous IR/PPP2R2A mice showed significantly impaired glucose tolerance compared with either mutation alone or wild-type littermates at 12 weeks of age (Fig. 2D). All three groups compared with the wild-type mice were insulin resistant (Fig. 2E); however, the doubly heterozygous IR/PPP2R2A mice showed significantly higher starting glucose levels compared with the other genotype classes (17.03 mmol/L vs. 10.89–12.63 mmol/L, respectively). A highly compensatory increase in pancreatic β-cell mass was observed in all three groups compared with wild-type littermates (Fig. 2F). No difference in body weight or food intake was observed in metabolic caging; however, double heterozygous IR/PPP2R2A mice drank more and produced more urine as glycosuria developed (Supplementary Fig. 3).

To investigate the effect of missplicing of the Ppp2r2a transcript on the PPP2R2A protein, we extracted protein from doubly heterozygous IR/PPP2R2A or heterozygous IR mice and found significantly reduced PPP2R2A in the liver (23% less), skeletal muscle (43%), and white adipose tissue (WAT) (22%) (Fig. 3A–C).

PPP2R2A has been shown to target the PP2A holoenzyme to AKT, which then selectively dephosphorylates Thr-308, returning AKT to the available protein pool (10). However, constitutively phosphorylated AKT is targeted to the proteasome for degradation (11). Consistent with the latter, we found a significant reduction in total AKT protein levels in the liver (25%), skeletal muscle (40%), and WAT (18%) in doubly heterozygous IR/PPP2R2A mice compared with heterozygous IR mice (Fig. 3A–C). Further, we found a significant reduction in the amount of insulin-stimulated Thr-308 phosphorylated AKT (normalized to tubulin to account for the differences in total AKT) in the liver, skeletal muscle, and WAT in doubly heterozygous IR/PPP2R2A mice that had been fasted overnight and given a bolus of either insulin or saline via the hepatic portal vein (Fig. 3D–F). Additionally, we also observed a reduction in Ser-473 phosphorylation of AKT, but this was only statistically significant in the liver (Fig. 3G–I).

Downstream AKT signaling was assessed by measuring the phosphorylation of both GSK-3β and p70S6K, total levels of which were not significantly changed (Fig. 4A–C). A significant reduction in the degree of insulin-stimulated GSK-3β and p70S6K phosphorylation was observed in all three tissues in doubly heterozygous IR/PPP2R2A mice compared with heterozygous IR mice (Fig. 4D–I).

In order to confirm that the Ppp2r2a mutation is hypomorphic, thus resulting in AKT dysregulation, we used siRNAs to knockdown Ppp2r2a expression in hepatocytes (Hepa1-6), myoblasts (C2C12), and adipocytes (3T3-L1). Following knockdown, cells were treated with either insulin or saline (basal), and then protein and RNA were extracted for analysis. 3T3-L1 cells were differentiated into adipocytes prior to siRNA knockdown on day 8 and treated with insulin or saline on day 10. Differentiation was confirmed by the accumulation of lipid droplets and Fabp4 gene expression (Supplementary Fig. 4). In all cell lines, siRNA treatment resulted in a significant reduction in Ppp2r2a mRNA levels (Supplementary Fig. 5).

Consistent with the in vivo tissue analysis, there was a significant reduction in total PPP2R2A protein levels (82.3% reduction Hepa1-6, 68.3% reduction C2C12, and 35.5% reduction 3T3-L1). This resulted in a significant reduction in AKT protein levels (61.2% reduction Hepa1-6, 21.8% reduction C2C12, and 26.5% reduction 3T3-L1) (Table 1).

In siRNA-treated Hepa1-6 cells, a small but significant increase in basal Thr-308 phosphorylated AKT (relative to tubulin) was observed and a significant reduction in insulin-stimulated Thr-308 and Ser-473 phosphorylated AKT in Ppp2r2a-treated compared with nonsense siRNA (control)–treated cells. Similarly, in C2C12 cells, there was a significant increase in basal and a reduction in insulin-stimulated Thr-308 and Ser-473 phosphorylated AKT. In 3T3-L1 cells, there was a significant increase in basal but no difference in insulin-stimulated Thr-308 and Ser-473 phosphorylated AKT in Ppp2r2a-silenced cells compared with control-treated cells (Table 1).

Next, we examined insulin-stimulated AKT signaling in Ppp2r2a-silenced cell lines. Under basal saline conditions, there was a higher proportion of GSK-3β phosphorylated in Hepa1-6 and 3T3-L1 Ppp2r2a-silenced cells compared with control-treated cells. However, the proportion of GSK-3β that was phosphorylated on insulin stimulation was less in Ppp2r2a-silenced cells, being largely unresponsive to the effect of insulin compared with control-treated cells in hepatocyte (Hepa1-6) and muscle (C2C12) but not in adipocyte (3T3-L1) cells (Table 2).

A significant increase in the proportion of basal phosphorylation of p70S6K was also observed in all three cell lines after Ppp2r2a knockdown compared with control-treated cells. The proportion of insulin-stimulated p70S6K phosphorylation was similar compared with control-treated cells, with a small but significant increase in total GSK-3β and p70S6K protein levels observed in silenced C2C12 cells (both basal and insulin treated) but not in the other cell lines (Table 2).

AKT phosphorylates and regulates forkhead box O1 (FOXO1), which regulates the transcription of gluconeogenic and lipogenic genes. Foxo1 gene expression is increased in the fasted state and is reduced after feeding (12). Therefore, we compared serum-fed and -fasted (18 h) siRNA-treated Hepa1-6 cells to test whether reduced AKT signaling led to altered regulation of FOXO1 targets. Instead of suppression, we found Foxo1 mRNA expression was elevated in Ppp2r2a-silenced, relative to control-treated, cells under serum treatment (Fig. 5A).

A significant increase in phosphoenolpyruvate carboxykinase (PEPCK [Pck1]), glucose-6-phosphatase catalytic subunit (G6pc), and Irs-2 gene expression was observed in Ppp2r2a siRNA-treated cells compared with control-treated cells under serum treatment (Fig. 5B–D). The transcription factor sterol regulatory element–binding protein transcription factor 1 (SREBF1) regulates genes required for glucose metabolism and fatty acid and lipid production, and we observed a significant reduction in both serum-fed and serum-free expression of Srebf1c (Fig. 5E) in Ppp2r2a siRNA-treated cells compared with control-treated cells. Most strikingly, Srebf1c levels were similarly low under both serum-fed and serum-free conditions in Ppp2r2a siRNA-treated cells. This resulted in a significant reduction in mRNA expression of fatty acid synthase (Fasn) and insulin-induced gene 1 (Insig1) targets of SREBF1 under both sets of conditions (Fig. 5F and G). Peroxisome proliferator–activated receptor γ, coactivator 1 α (Ppargc1a), a transcriptional coactivator that interacts with and regulates the activities of the cAMP-responsive element–binding protein to drive the transcription of gluconeogenic genes, was also significantly upregulated with serum treatment in Ppp2r2a-silenced cells compared with control-treated cells (Fig. 5H). Glucose production assays were performed to measure the physiological effect of altered gluconeogenesis gene transcription in siRNA-silenced cells. As predicted, control-treated cells produced significantly less glucose when treated with insulin; however, Ppp2r2a siRNA-treated cells still secreted significant levels of glucose into the media despite insulin treatment (Fig. 5I).

For comparison with the in vivo state, expression analysis was repeated in fed compared with fasted liver cDNA from doubly heterozygous IR/PPP2R2A and heterozygous IR mice. As observed in silenced cells, Foxo1 expression was increased (Fig. 6A), with a concomitant significant increase in Pck1, G6pc, and Irs-2 gene expression (Fig. 6B–D) and a decrease in Srebf1c and Fasn (Fig. 6E and F) in doubly heterozygous IR/PPP2R2A fed mice. Levels of Fasn and Insig1 were also significantly reduced in fasted doubly heterozygous IR/PPP2R2A compared with heterozygous IR mice (Fig. 6F and G). Ppargc1a was upregulated in doubly heterozygous IR/PPP2R2A compared with heterozygous IR fed mice (Fig. 6H).

We have identified a point mutation in intron 3 of the Ppp2r2a gene that resulted in modestly increased fasted insulin and insulin resistance, as did an IR heterozygous knockout mutation. Both mutations together in doubly heterozygous IR/PPP2R2A mice resulted in diabetes—significant hyperglycemia, hyperinsulinemia, impaired glucose tolerance, and glycosuria. This additive digenic effect of two mutations is similar to mice doubly heterozygous for null alleles of both IR and IRS-1, components of the insulin signaling pathway, which develop diabetes (4).

The splice mutation reduced Ppp2r2a mRNA levels, although some normal splicing was still retained from the mutant allele, and reduced levels of PPP2R2A protein showed the hypomorphic nature of the mutation. Knockdown studies using siRNAs in insulin-responsive cell lines support the hypothesis that the reduction in PPP2R2A protein levels rather than the production of a truncated dominant-negative protein product is responsible for the phenotype. The PP2A holoenzyme consists of a dimeric core enzyme that is composed of a catalytic subunit C, a structural subunit A, and a variable regulatory subunit B. There are at least four subfamilies of regulatory B subunits (including PPP2R2A), and it is thought that this diversity in regulatory subunits dictates substrate specificity and the subcellular localization of PP2A (13). PPP2R2A is expressed in a wide range of tissues including, but not limited to, insulin-sensitive tissues (liver, skeletal muscle, and adipose tissue) and β-cells of the pancreatic islet (14–16). PP2A is a tumor suppressor and is inactivated or downregulated in colorectal cancer, myeloid leukemia, small-cell lung carcinomas, and luminal breast cancers (17–22). The role of PPP2R2A in cell growth and division may explain its role as a tumor suppressor; however, milder hypomorphic alleles may modulate insulin signaling, in conjunction with other defects, without such a clear alteration of cancer risk.

AKT, a key node in insulin signaling, is phosphorylated by 3-phosphoinositide–dependent kinase 1 at Thr-308 and a mechanistic target of rapamycin (mTOR) at Ser-473 in response to an insulin signal, whereas PPP2R2A containing PP2A holoenzyme specifically dephosphorylate AKT at Thr-308 but not Ser-473 (5,10). Reduced dephosphorylation of AKT due to reduced PPP2R2A containing PP2A may result in AKT association with the COOH terminus of Hsp70-interacting protein (CHIP) ubiquitination and then degradation (11). This could explain the decrease in AKT protein levels that we observed in doubly heterozygous IR/PPP2R2A mice and Pppr2r2a siRNA-silenced cell lines. PPP2R2A levels were reduced to lower levels in siRNA-treated cells than observed in mouse tissues, resulting in a more severe reduction in AKT protein levels and consequently less insulin-induced phosphorylation of AKT and its protein targets. Strikingly, in Pppr2r2a siRNA-silenced cells, we observed significantly increased basal (saline treated, serum starved) levels of AKT phosphorylation compared with control-treated cells, suggesting a significant reduction in PP2A dephosphorylation of AKT in acutely silenced cells. This prolonged active basal AKT state is also reflected in the cell line–specific increase in basal phosphorylation of GSK-3β and p70S6K.

Reduced suppression of hepatic glucose output (HGO) is a hallmark of type 2 diabetes, and HGO is regulated through the activation of AKT and phosphorylation of FOXO1 (23–26). Phosphorylation of FOXO1 results in its translocation out of the nucleus, reducing gluconeogenic gene expression (24,25,27,28) and promoting cytoplasmic ubiquitination and degradation (29–31). PPP2R2A containing PP2A holoenzyme has also been shown to specifically dephosphorylate FOXO1 in islet β-cells under oxidative stress (15). The expression of Foxo1 RNA was increased in serum-fed Ppp2r2a-silenced cells compared with control-treated Hepa1-6 cells and was similarly increased in fed doubly heterozygous IR/PPP2R2A compared with heterozygous IR mice. In contrast, in control-treated cells, the suppression of Foxo1 RNA expression was seen in both serum-treated (compared with serum-free–treated) cells and fed (compared with fasted) liver tissue. Consistent with elevated Foxo1 expression in mutant mice or cells, there was increased RNA expression of its gluconeogenic target genes Pck1 (Pepck), Ppargc1a (Pgc1a), and G6pc and another Foxo1 target Irs-2 in both cells and tissues. This could lead to a failure of insulin to suppress HGO and likely explains a large proportion of the observed phenotype of the mice.

p70S6K is activated in the insulin signaling pathway via AKT phosphorylation of mTOR, which in turn phosphorylates and activates p70S6K. A well-defined negative signaling pathway has been described involving negative phosphorylation of IRS-1 by p70S6K, leading to insulin-induced degradation of IRS-1 (32,33). Insulin-stimulated p70S6K phosphorylation levels were significantly reduced in the liver, skeletal muscle, and WAT in doubly heterozygous IR/PPP2R2A mice. However, in Ppp2r2a siRNA-silenced cell lines, which showed a greater reduction in PPP2R2A protein levels than observed in IGT10 mice, a significant increase in basal p70S6K phosphorylation was seen, which could result in reduced IRS-1 activity. This was not apparent in insulin-treated cells where levels were similar to control-treated cells. Therefore, although basal dysregulation could contribute to insulin resistance at least in cell lines, where knockdown is acute, it seems unlikely to be the primary explanation for the in vivo phenotypes.

The insulin-regulated activity of AKT regulates SREBP1c activity, which is the master regulator of the transcription of key lipogenic genes (6). Consistent with reduced insulin-stimulated AKT signaling, Srebf1c expression was reduced in serum-fed Ppp2r2a-silenced cells compared with control-treated Hepa1-6 cells and was similarly decreased in fed doubly heterozygous IR/PPP2R2A compared with heterozygous IR mice. Consequently, Fasn and Insig1, two SREBP1c-regulated genes, were also downregulated under these conditions, therefore disrupting lipogenesis under fed conditions.

We found a reduced proportion of phosphorylated GSK-3β in response to insulin stimulation in tissues from doubly heterozygous IR/PPP2R2A compared with heterozygous IR mice and in corresponding Hepa1-6 and C2C12 Ppp2r2a-silenced cells compared with control-treated cells, clearly showing impaired insulin signaling. As unphosphorylated active GSK-3β is a negative regulator of glycogen synthase, there is likely to be less glycogen storage of glucose, increasing the supply of glucose into the blood and thus contributing to the phenotype (34).

Although PPP2R2A has not previously been linked to type 2 diabetes in humans, PP1A proteins have also been shown to have an important role in glucose metabolism via its regulatory effects on glycogen metabolizing enzymes, including glycogen synthase (GS), glycogen phosphorylase (GP), and GP kinase. Heterozygous knockout of PPP1R3C results in the reduction of glycogen stores, progressive glucose intolerance, hyperinsulinemia, and insulin resistance (35). A PPP1R3A regulatory subunit of PP1, which binds to muscle glycogen, enhances the dephosphorylation of glycogen-bound substrates for PP1, such as GS and GP kinase, which plays an important role in glycogen synthesis but is not essential for insulin activation of GS (36). Recently, the PP2A subunit PPP2R5C has been shown to couple hepatic glucose and lipid homeostasis with hepatic knockdown in mice, resulting in improved systemic glucose tolerance and insulin sensitivity (37). Such evidence suggests that negative control in the insulin signaling pathway by PP1A and PP2A may prove to be important in the susceptibility to type 2 diabetes in humans. Further, population-based genome-wide association studies show that the alteration of gene expression of genome-wide association study genes in specific tissues leads to increased diabetes risk, and our model is similar in that a hypomorphic allele of a PPP2R2A leads to diabetes in conjunction with insulin resistance.

Acknowledgments. The authors would like to thank the Mary Lyon Centre (Harwell, U.K.) for excellent mouse husbandry.

Funding. M.G., Y.B., H.H., C.T.E., and R.D.C. were funded by the Medical Research Council (MC_U142661184).

Duality of Interest. The gene identification aspect of this research was a collaboration between Amgen Inc. and Medical Research Council Harwell. Amgen Inc. also provided funding to R.D.C. for these activities. C.-M.L., H.G., E.L., D.B., W.B., T.J., M.M.V., and D.J.L. are employees of Amgen Inc. and have no conflict/duality of interest. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.G. contributed to the design and carried out the experimental work and preparation of the manuscript. Y.B. contributed to the technical setup of the siRNA experiments. C.-M.L. carried out the DNA-Seq library construction and next-generation sequencing (NGS). H.G. carried out the NGS data analysis and identification of ENU candidate hits. E.L. and D.B. carried out all the subsequent genotyping for Ppp2r2a and Itm2b mutations. H.H. and C.T.E. ran the samples for the PEGGY experiments. W.B. validated the ENU candidates via Sanger sequencing. T.J. was the team leader overseeing the ENU/NGS efforts at Amgen Inc. M.M.V., D.J.L., and R.D.C. contributed to experimental design and manuscript preparation. M.G. and R.D.C. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.