Protein-tyrosine Phosphatase 1B Deficiency Reduces Insulin Resistance and the Diabetic Phenotype in Mice with Polygenic Insulin Resistance*

Mice heterozygous for insulin receptor (IR) and IR substrate (IRS)-1 deficiency provide a model of polygenic type 2 diabetes in which early-onset, genetically programmed insulin resistance leads to diabetes. Protein-tyrosine phosphatase 1B (PTP1B) dephosphorylates tyrosine residues in IR and possibly IRS proteins, thereby inhibiting insulin signaling. Mice lacking PTP1B are lean and have increased insulin sensitivity. To determine whether PTP1B can modify polygenic insulin resistance, we crossed PTP1B–/– mice with mice with a double heterozygous deficiency of IR and IRS-1 alleles (DHet). DHet mice weighed slightly less than wild-type mice and exhibited severe insulin resistance and hyperglycemia, with ∼35% of DHet males developing diabetes by 9–10 weeks of age. Body weight in DHet mice with PTP1B deficiency was similar to that in DHet mice. However, absence of PTP1B in DHet mice markedly improved glucose tolerance and insulin sensitivity at 10–11 weeks of age and reduced the incidence of diabetes and hyperplastic pancreatic islets at 6 months of age. Insulin-stimulated phosphorylation of IR, IRS proteins, Akt/protein kinase B, glycogen synthase kinase 3β, and p70S6K was impaired in DHet mouse muscle and liver and was differentially improved by PTP1B deficiency. In addition, increased phosphoenolpyruvate carboxykinase expression in DHet mouse liver was reversed by PTP1B deficiency. In summary, PTP1B deficiency reduces insulin resistance and hyperglycemia without altering body weight in a model of polygenic type 2 diabetes. Thus, even in the setting of high genetic risk for diabetes, reducing PTP1B is partially protective, further demonstrating its attractiveness as a target for prevention and treatment of type 2 diabetes.

Insulin plays a dominant role in regulating glucose homeostasis through a highly orchestrated constellation of effects, which include promoting glucose uptake in peripheral tissues such as muscle and fat, suppressing hepatic glucose output, and regulating lipid metabolism. Failure of peripheral tissues to respond to insulin (i.e. insulin resistance) can initially be overcome by a compensatory increase in insulin secretion from pancreatic ␤-cells. When this mechanism fails to compensate sufficiently, frank diabetes and its attendant complications ensue (1)(2)(3)(4).
Insulin action is mediated through a complex network of signaling events, which are initiated by the binding of insulin to its cell-surface receptor, the insulin receptor (IR). 3 This triggers the intrinsic protein-tyrosine kinase activity of IR, resulting in autophosphorylation of several IR tyrosyl residues and the recruitment and tyrosyl phosphorylation of IR substrate (IRS) proteins. Subsequently, molecules such as the growth factor receptor-binding protein Grb-2, the p85 regulatory subunit of phosphatidylinositol 3-kinase (PI3K), and the SH2 domaincontaining protein-tyrosine phosphatase Shp2 bind to IRS proteins, leading to the metabolic and growth-promoting effects of insulin (2,3).
The insulin signaling cascade is negatively regulated by protein-tyrosine phosphatases, most notably protein-tyrosine phosphatase 1B (PTP1B). PTP1B, which is widely expressed and localizes predominantly to the endoplasmic reticulum (5,6), is a major IR phosphatase (7,8), dephosphorylates IR preferentially at the tandem Tyr 1162 and Tyr 1163 residues in vitro (9) and in cultured cells (10), and may also dephosphorylate IRS-1 (11,12). Its expression and activity are increased in obese and insulin-resistant human subjects as well as in some obese rodent models (13)(14)(15)(16), and some studies suggest that PTP1B polymorphisms may be associated with obesity and insulin resistance in humans (17)(18)(19). The overexpression of PTP1B in muscle results in insulin resistance (20). Conversely, mice lacking PTP1B have markedly increased insulin sensitivity (21,22) associated with increased or prolonged insulin-stimulated IR and IRS-1 tyrosine phosphorylation in muscle and liver (21).
The absence of PTP1B in all tissues is also associated with leanness and protection from diet-induced obesity owing in large part to enhanced leptin signaling because PTP1B also dephosphorylates Jak2 in the leptin signaling cascade (23,24). In general, increased adiposity is a major cause of insulin resistance (25). Therefore, it has been difficult to determine the extent to which enhanced insulin sensitivity in PTP1B Ϫ/Ϫ mice is due to the direct effect of the absence of PTP1B on IR signaling, as opposed to more indirect effects resulting from leanness. We recently found that deletion of PTP1B selectively in neurons increases insulin sensitivity, but as in total body PTP1B Ϫ/Ϫ mice, adiposity is reduced in neuron-specific PTP1B Ϫ/Ϫ mice (26). Reduction of PTP1B expression primarily in liver and adipose tissue of ob/ob mice by antisense oligonucleotides has also been reported to improve insulin sensitivity, but again, largely in association with decreased adiposity (27)(28)(29). All of these data strongly support a physiological role for PTP1B in negatively regulating insulin signaling in rodents and humans. However, the conclusions are tempered by the fact that the accompanying changes in adiposity may play a major role in altering insulin signaling and systemic insulin sensitivity.
To determine the role of PTP1B in modulating insulin signaling independent of changes in adiposity in a highly relevant diabetes model, we crossed homozygous PTP1B Ϫ/Ϫ mice with mice with a double heterozygous deficiency of IR and IRS-1 alleles (DHet) (30). The DHet model has similarities to human type 2 diabetes in that it is polygenic in nature, has early-onset insulin resistance and is modified by both genetic background and diet (30,31). These mice have a 35-65% reduction in the levels of IR and IRS-1 in insulin target tissues. Although there is no apparent phenotype in either IR ϩ/Ϫ or IRS-1 ϩ/Ϫ mice, the combination of these allelic deletions results in severe hyperinsulinemia, insulin resistance, ␤-cell hyperplasia, and ultimately, frank diabetes in 50% (30) to 90% (this study) of the mice on a mixed genetic background.
In this study, we show that PTP1B deficiency reduces the severe hyperinsulinemia, insulin resistance, hyperglycemia, and ␤-cell hypertrophy in DHet mice without affecting their body weight. Thus, we demonstrate that the absence of PTP1B protects against diabetes: 1) even in mice with a high risk for diabetes due to reduction of key signaling molecules and 2) in the absence of alterations in body weight. These data support the notion that, even in the setting of genetic defects that markedly increase diabetes risk, PTP1B is an attractive drug target for the prevention and treatment of type 2 diabetes.
Metabolic Measurements-Blood was collected via tail bleed in mice in the fed state between 8 and 10 a.m. Blood glucose was measured using a OneTouch Ultra glucose meter (LifeScan, Inc., Milpitas, CA). Serum insulin and leptin levels were measured by enzyme-linked immunosorbent assay (Crystal Chem Inc., Chicago, IL). Serum-free fatty acid levels in mice in the fed state were measured using a kit from Wako Pure Chemical Industries (Richmond, VA). For glucose tolerance tests, mice were fasted overnight, and blood glucose was measured immediately before and 15, 30, 60, 90, and 120 min after intraperitoneal injection of glucose (1.5 g/kg of body weight). For insulin tolerance tests, food was removed for 4 h, and blood glucose was measured immediately before and 15, 30, 60, 90, and 120 min after intraperitoneal injection of human insulin (1.2 and 1.0 unit/kg of body weight for males and females, respectively; Humulin, Lilly). All metabolic measurements were performed when mice were 9 -11 weeks and 6 months of age.
Insulin Signaling Studies-Six-month-old male WT, PTP1B Ϫ/Ϫ , DHet, and DHet/PTP1B Ϫ/Ϫ mice were fasted overnight. Human insulin (10 units/kg of body weight; Humulin) was injected via tail vein; 7.5 min later, mice were killed by CO 2 , and tissues were quickly collected and snap-frozen in liquid nitrogen. Tissues were stored at Ϫ80°C until processing. This time point was chosen to optimize the possibility of seeing increased or prolonged signaling in both muscle and liver in the absence of PTP1B (21). In our studies in normal mice, maximal responses have been seen with this insulin dose. Blood glucose levels were between 300 and 500 mg/dl in the majority of male DHet mice and Ͻ200 mg/dl in the majority of male DHet/ PTP1B Ϫ/Ϫ mice at 6 months of age. Therefore, we chose mice with these blood glucose levels for the insulin signaling studies because we wanted to investigate whether the improved insulin sensitivity in DHet/PTP1B Ϫ/Ϫ mice could be explained by their improved insulin signaling.
PI3K Activity-Muscle and liver lysates (500 g of protein) were subjected to immunoprecipitation overnight at 4°C with 3 l of either anti-IRS-1 or anti-IRS-2 antibody coupled to protein A-Sepharose (Sigma). The immune complex was washed, and PI3K activity was determined as described previously (34).
RNA Extraction and Real-time PCR-RNA was extracted from liver using TriReagent (Molecular Research Center, Inc., Cincinnati, OH) and quantitated by measuring the absorbance at 260 nm. The mRNA level of sterol regulatory element-binding protein 1c (SREBP1c) was quantified by quantitative real-time reverse transcription-PCR as described previously (54). The reverse transcription-PCR conditions were 48°C for 30 min for reverse transcription, followed by 95°C for 10 min and then 40 cycles at 95°C for 15 s and 60°C for 1 min. The relative copy number of cyclophilin was quantified and used for normalization. The mRNA level was calculated using a standard curve with a series of 10-fold dilutions of a mixture of representative RNA samples. Primer and probe sequences were as follows: SREBP1c, 5Ј-GACCACGGAGCCATGGA-3Ј (forward), 5Ј-GGGAAGTCACTGTCTTGGTTGTT-3Ј (reverse), and 5Ј-TGCACATTTGAAGACATGCTCCAGCTCAT-3Ј (probe); and cyclophilin, 5Ј-GGTGGAGAGCACCAAGACAGA-3Ј (forward), 5Ј-GCCGGAGTCGACAATGATG-3Ј (reverse), and 5Ј-TCCTT-CAGTGGCTTGTCCCGGCT-3Ј (probe).
Morphological and Morphometric Analysis of Pancreatic Islets-Male mice were killed at 6 months of age by CO 2 , and pancreata were removed, cleared of fat and lymph nodes, weighed, spread in anatomical orientation, and fixed in either 10% formalin or Bouin's solution. Full footprint paraffin sections (5 m) were immunostained for ␤-cells using a guinea pig anti-porcine insulin antiserum (Linco, St. Louis, MO) as described previously (35). ␤-Cell mass was measured by point-counting morphometry (35). Islet morphology and morphometry were assessed without prior knowledge of the genotype.
Statistical Analysis-Results are presented as the mean Ϯ S.E. Differences between groups were analyzed for statistical significance by analysis of variance (ANOVA) with Fischer's probable least-squares difference post hoc test, ANOVA with repeated measures, or Kruskal-Wallis non-parametric ANOVA by rank as appropriate.

RESULTS
Mice of the four genotypes WT, PTP1B Ϫ/Ϫ , DHet, and DHet/PTP1B Ϫ/Ϫ were used in this study. There was no difference in body weight between WT and PTP1B Ϫ/Ϫ mice on a standard chow diet at 5 weeks of age, but by 11 weeks (males) or 15 weeks (females), PTP1B Ϫ/Ϫ mice weighed less than WT mice ( Fig. 1). PTP1B Ϫ/Ϫ mice also had lower epididymal fat pad mass (Table 1), indicating lower adiposity, as shown previously (22). Although liver mass was lower in PTP1B Ϫ/Ϫ mice (Table  1), the effect was no longer present when corrected for body weight (data not shown). Male and female DHet mice weighed less than their WT littermates, as described previously (30,31), and had lower fat pad mass even when corrected for lower body weight, i.e. the fat pad/body weight ratio was lower (Table 1). This most likely reflects insulin resistance in adipose tissue (31). Liver mass in DHet mice tended to be lower than that in WT   AUGUST 17, 2007 • VOLUME 282 • NUMBER 33

JOURNAL OF BIOLOGICAL CHEMISTRY 23831
mice, but not after correction for lower body weight (Table 1). Superimposition of PTP1B deficiency on the DHet mice (DHet/PTP1B Ϫ/Ϫ ) had no further effect on body weight ( Fig. 1), fat pad or liver mass, or fat pad/body weight ratio compared with DHet mice with intact PTP1B (Table 1). At 9 -10 weeks of age, male PTP1B Ϫ/Ϫ mice had lower serum insulin and blood glucose levels compared with WT mice (p Ͻ 0.05 versus WT mice) (Fig. 2, A and B), consistent with the higher insulin sensitivity in PTP1B Ϫ/Ϫ mice observed previously (21,22). The glucose ϫ insulin product was also significantly 2C). There was no difference in fed insulin or glucose levels between female PTP1B Ϫ/Ϫ and WT mice (Fig. 2, D and E), similar to our previous results (22), although the glucose ϫ insulin product tended to be lower in female PTP1B Ϫ/Ϫ mice compared with WT mice (Fig. 2F). Because of their genetically determined insulin resistance, most male DHet mice developed severe hyperinsulinemia in the fed state, with fed insulin levels ranging 2.6 -20.5-fold above WT levels ( Fig. 2A). Of 31 male DHet mice examined at 9 -10 weeks of age, 11 mice (35%) developed severe fed hyperglycemia of Ͼ250 mg/dl, whereas the remaining 65% had normal glycemia ( Fig.  2B), consistent with the heterogeneity of diabetes incidence in this model (30). A blood glucose value of 250 mg/dl was used as the cutoff because no male WT mice had a higher blood glucose level. The degree of hyperinsulinemia and hyperglycemia was milder in female DHet mice (Fig. 2, D and E). When insulin sensitivity was expressed as the glucose ϫ insulin product, both male and female DHet mice displayed extremely high values (20and 8-fold above WT levels, respectively; p Ͻ 0.05) (Fig. 2, C and F). PTP1B deficiency completely normalized blood glucose levels in mice in the fed state (Fig. 2, B and E) and partially rescued hyperinsulinemia in male and female DHet/PTP1B Ϫ/Ϫ mice (Fig. 2, A and D). PTP1B deficiency in DHet mice also greatly reduced the glucose ϫ insulin product (Fig. 2, C and F). Heterogeneity in blood glucose and insulin levels is found in both DHet and DHet/ PTP1B Ϫ/Ϫ mice, most likely reflecting genetic modifiers due to their mixed genetic background (30,31). Different degrees of severity of insulin resistance have previously been reported in DHet mice on different genetic backgrounds (31).
DHet mice exhibited impaired glucose tolerance and insulin resistance as determined by intraperitoneal glucose and insulin tolerance tests, respectively. Glucose tolerance was enhanced in male PTP1B Ϫ/Ϫ mice compared with WT mice (Fig. 3A), as shown previously (21,22). The absence of PTP1B in DHet mice  Serum-free fatty acid levels were lower in 10-week-old male PTP1B Ϫ/Ϫ and DHet mice in the fed state compared with WT mice (supplemental Fig. 2). This most likely reflects the reduced adipose mass in these genotypes. The difference in serum-free fatty acids in DHet mice was more significant on this mixed genetic background (supplemental Fig. 2) than reported previously for inbred DHet mice (31). Serum-free fatty acid levels were even lower in DHet/PTP1B Ϫ/Ϫ mice compared with DHet mice (supplemental Fig. 2), which could indicate improved insulin sensitivity.
As expected from our previous study, serum leptin levels were lower in 10-week-old male PTP1B Ϫ/Ϫ mice compared with WT mice (supplemental Fig. 3), consistent with improved leptin sensitivity (22). Serum leptin levels tended to be lower in DHet mice compared with WT mice (supplemental Fig. 3) although this did not reach statistical significance even when expressed as a ratio of leptin to body weight (data not shown). Previous data showed that serum leptin levels are lower in DHet mice on inbred backgrounds compared with WT mice (31), most likely reflecting their reduced fat mass. Leptin levels in DHet/PTP1B Ϫ/Ϫ mice were similar to those in DHet mice (supplemental Fig. 3).
Of 35 WT males examined at 6 months of age, 23% developed hyperinsulinemia of Ͼ5 ng/ml (Fig. 4A), and 8% developed hyperglycemia of Ͼ250 mg/dl (Fig.  4B). A previous study has shown that, with aging and depending on their genetic background, WT mice can develop hyperinsulinemia and hyperglycemia at frequencies similar to those observed herein (31). In contrast, serum insulin and glucose levels remained normal in all PTP1B Ϫ/Ϫ mice, demonstrating a protective role for PTP1B deficiency against the insulin resistance and diabetes associated with aging. The improved insulin sensitivity was also reflected in the glucose ϫ insulin product, which increased by ϳ7-fold in WT mice between 2 and 6 months of age, but only by ϳ3-fold in PTP1B Ϫ/Ϫ mice (Figs. 2C and 4C).
Of the 38 female DHet mice examined, 61% developed hyperinsulinemia (defined by a serum insulin value of 2.5 ng/ml because no female WT mice had insulin levels higher than this value). Twenty-eight percent of DHet females developed hyperglycemia (defined by a glucose value of 200 mg/dl because no WT females had blood glucose levels higher than this value). However, only 5% of female DHet/PTP1B Ϫ/Ϫ mice developed hyperinsulinemia of Ͼ2.5 ng/ml, and none developed hyperglycemia of Ͼ200 mg/dl (Fig. 4, D and E). Taken together, these data indicate that PTP1B deficiency can partially protect DHet mice from developing severe insulin resistance and diabetes even with aging. At all ages examined, the insulin resistance phenotype seems milder in female than male DHet mice, and the improvement of insulin sensitivity is greater in female than male DHet/PTP1B Ϫ/Ϫ mice. Sexual dimorphism in the extent of changes in insulin sensitivity in DHet (30) and PTP1B Ϫ/Ϫ (22) mice has previously been reported and has been seen in other models of type 2 diabetes (36) and improved insulin sensitivity (37).
Pancreatic sections from 6-month-old DHet mice showed enormously enlarged insulin-staining islets (Fig. 5A), as expected for severely insulin-resistant mice (30). The ␤-cell mass, determined by point-counting morphometry (35), was increased by 3-10-fold in DHet mice (Fig. 5B), which is comparable with the values reported previously (30). Serum insulin levels correlated well with glucose levels in DHet mice, i.e. higher insulin levels were found in mice with more severe degrees of hyperglycemia (Fig. 5C), similar to a previous description (30). This suggests that insulin secretion is increased in response to insulin resistance in this model, as seen in humans with type 2 diabetes. However, even with hyperplastic islets and markedly increased serum insulin levels, hyperglycemia occurred in the majority of DHet mice due to relative ␤-cell "insufficiency" in the setting of extreme insulin resistance.
We evaluated the expression levels of key molecules in the insulin signaling pathway. Deletion of PTP1B did not alter IR or IRS-1 levels in muscle or liver compared with WT mice (supplemental Fig. 4). As expected, IR levels were reduced by 35-40% in muscle and liver of DHet and DHet/PTP1B Ϫ/Ϫ mice (p Ͻ 0.05). IRS-1 levels were reduced by ϳ65% in muscle and by ϳ50% in liver of DHet and DHet/PTP1B Ϫ/Ϫ mice (p Ͻ 0.05 for both tissues). IRS-2, Akt/PKB, GSK3␤, and p70 S6K levels were not different among the four genotypes (supplemental Fig. 4).
To assess the consequence of PTP1B deficiency in early steps in insulin signaling in DHet mice, we evaluated tyrosyl phosphorylation sites critical for activation of the IR signaling cascade, including autophosphorylation of IR at (i) the tandem tyrosyl residues Tyr 1162 and Tyr 1163 , located in the activation loop of the IR kinase domain and required for IR activation and subsequent phosphorylation of other IR tyrosines (39); and (ii) the juxtamembrane residue Tyr 972 , which contributes to IRS-1 recruitment (40). We also measured phosphorylation of IRS-1 at Tyr 612 , which is important for full activation of PI3K and subsequent translocation of Glut4 in response to insulin (41).
In muscle, basal (saline-injected) IR, IRS-1, and Akt/PKB phosphorylation levels were very low in all four genotypes (Fig.  6). In WT mice, insulin rapidly stimulated IR phosphorylation at Tyr 1162 /Tyr 1163 and Tyr 972 and IRS-1 phosphorylation at Tyr 612 (Fig. 6, A-C). Tyrosyl phosphorylation of IR and IRS-1 was further enhanced in PTP1B Ϫ/Ϫ mice, consistent with their enhanced insulin sensitivity (21,22). Phosphorylation of IR at Tyr 1162 /Tyr 1163 and Tyr 972 was increased by 30 -40% (p Ͻ 0.05 versus WT mice) (Fig. 6, A and B). Phosphorylation of IRS-1 at Tyr 612 was increased by ϳ150% versus WT mice (Fig. 6C). In muscle of DHet mice, IR Tyr 1162 /Tyr 1163 and Tyr 972 phosphorylation was reduced by ϳ70% versus WT mice, and there was an ϳ60% reduction of IRS-1 Tyr 612 phosphorylation. Lack of FIGURE 5. A, pancreatic islet morphology in male WT, PTP1B Ϫ/Ϫ , DHet, and DHet/PTP1B Ϫ/Ϫ mice at 6 months of age. Pancreata were removed, fixed in 10% formalin, paraffin-embedded, and sectioned as a full footprint. Tissue sections were immunostained with anti-insulin antibody and counterstained with hematoxylin. B, ␤-cell mass/pancreas in DHet and DHet/PTP1B Ϫ/Ϫ mice. The dotted line represents values for WT mice of the same gender and age and similar genetic background determined previously (30). C, correlation between fed insulin and glucose levels in male DHet mice. *, p Ͻ 0.05 versus DHet mice.
When IR and IRS-1 tyrosyl phosphorylation levels were corrected for total IR and IRS-1 protein levels, phosphorylation of IR at Tyr 1162 /Tyr 1163 and Tyr 972 and that of IRS-1 at Tyr 612 were still elevated in muscle of PTP1B Ϫ/Ϫ mice compared with WT mice (Fig. 6, A-C), suggesting that the stoichiometry of phosphorylation at these sites is increased in PTP1B Ϫ/Ϫ mice. Even after correcting for total IR protein levels, phosphoryla-tion of IR at Tyr 1162 /Tyr 1163 (Fig.  6A) and Tyr 972 (Fig. 6B) was still lower in muscle of DHet mice compared with WT mice, whereas phosphorylation of IRS-1 at Tyr 612 was similar in muscle of DHet and WT mice (Fig. 6C). These data suggest that heterozygous deficiency of IR and IRS-1 differentially affects the stoichiometry of phosphorylation of IR and IRS-1 at these important sites. Lack of PTP1B in DHet mice increased phosphorylation of IR at Tyr 1162 /Tyr 1163 and Tyr 972 above the levels in DHet mice, but the levels were still lower than those in WT mice (Fig. 6, A and B). When corrected for total IRS-1 protein levels, PTP1B deficiency resulted in an increased stoichiometry of IRS-1 phosphorylation at Tyr 612 in muscle of DHet/PTP1B Ϫ/Ϫ mice above the levels in WT mice and similar to the levels in PTP1B Ϫ/Ϫ mice (Fig. 6C).
In muscle of WT mice, insulin stimulation resulted in a profound increase in Akt/PKB Ser 473 phosphorylation, and PTP1B deficiency led to a further 50% enhancement of Akt/PKB phosphorylation (Fig. 6E). Heterozygous deficiency of IR and IRS-1 resulted in a 35% reduction of Akt/PKB phosphorylation in response to insulin stimulation, and lack of PTP1B in DHet mice increased Akt/PKB phosphorylation to the level observed in PTP1B Ϫ/Ϫ mice (1.2-fold (p Ͻ 0.05) and 45% (p Ͻ 0.05) increases compared with DHet and WT mice, respectively) (Fig. 6E). Thus, partial rescue of IR tyrosyl phosphorylation and full restoration of tyrosyl-phosphorylated IRS-1 result in further enhancement of phosphorylation of the downstream target, Akt/PKB, to the levels seen in PTP1B Ϫ/Ϫ mice. This indicates that there may be novel PTP1B targets in the insulin signaling pathway distal to IRS-1-associated PI3K activity that also modulate Akt/PKB activity. Insulin also stimulated the rapid phosphorylation of GSK3␤ and p70 S6K in muscle of WT mice, but there was no difference in phosphorylation of these substrates in mice of the different genotypes (Fig. 6, F  and G).  (IR pY972; B), IRS-1 Tyr 612 (IRS-1 pY612; C), Akt/PKB Ser 473 (E), GSK3␤ Ser 21 /Ser 9 (F), and p70 S6K Thr 389 (G) in muscle was measured by immunoblot analysis using antibodies specific for each phosphorylation site. The immunoblots shown are representative of three blots for each panel. In A-C, the upper bar graphs show total tyrosyl phosphorylation of IR and IRS-1 at specific sites, and the lower bar graphs show the specific tyrosyl phosphorylation corrected for IR and IRS-1 protein levels. IRS-1-associated PI3K activity was measured as described under "Materials and Methods" (D). For each genotype, n ϭ 3 for saline and n ϭ 6 -7 for insulin. *, p Ͻ 0.05 versus all other insulin-stimulated groups; #, p Ͻ 0.05 versus insulin-stimulated PTP1B Ϫ/Ϫ and DHet mice; ϩ, p Ͻ 0.05 versus insulin-stimulated WT and DHet mice as determined by ANOVA.
When corrected for total IR and IRS-1 protein levels, IR phosphorylation at Tyr 1162 /Tyr 1163 and Tyr 972 was still higher in liver of PTP1B Ϫ/Ϫ mice compared with WT mice (Fig. 7, A-C), suggesting increased stoichiometry of phosphorylation at these sites in PTP1B Ϫ/Ϫ mice. The stoichiometry of phosphorylation of IR Tyr 1162 /Tyr 1163 and Tyr 972 and IRS-1 Tyr 612 was similar in WT and DHet mice, whereas PTP1B deficiency increased the stoichiometry of phosphorylation of IR Tyr 1162 / Tyr 1163 to the WT level (Fig. 7A) and that of IR Tyr 972 (Fig. 7B) and IRS-1 Tyr 612 above the WT level (Fig. 7C).
In addition, insulin stimulated Akt/PKB Ser 473 phosphorylation in WT mice (Fig. 7G), and PTP1B deficiency led to a further enhancement of Akt/PKB Ser 473 phosphorylation (81% increase versus WT mice) (Fig. 7G). Heterozygous deficiency of IR and IRS-1 in DHet mice resulted in a 65% reduction in insulin-stimulated Akt/PKB phosphorylation compared with WT mice, whereas lack of PTP1B in DHet mice completely restored Akt/PKB phosphorylation to the levels observed in WT mice (Fig. 7G).
As in muscle, insulin stimulated phosphorylation of GSK3␤ in liver of WT mice, and GSK3␤ phosphorylation was increased by ϳ70% in PTP1B Ϫ/Ϫ mice (Fig. 7H). Heterozygous deficiency of IR and IRS-1 resulted in an ϳ40% reduction of GSK3␤ phosphorylation in liver (p Ͻ 0.05 versus WT mice), which was restored to WT levels in DHet/PTP1B Ϫ/Ϫ mice (Fig. 7H). Insulin stimulated phosphorylation of p70 S6K in WT and PTP1B Ϫ/Ϫ mice to similar extents (Fig. 7I). The level of insulin-stimulated p70 S6K phosphorylation was reduced by 38% in DHet mice and was restored to WT levels in DHet/PTP1B Ϫ/Ϫ mice (Fig. 7I). Thus, our data suggest that there are differential effects of PTP1B deficiency on downstream components of the insulin signaling pathway and that some of these effects are tissuespecific. In addition, expression of the key gluconeogenic enzyme phosphoenolpyruvate carboxykinase was elevated in liver of DHet mice and was restored to WT levels in DHet/ PTP1B Ϫ/Ϫ mice (Fig. 7J).
PTP1B induces the expression of SREBP1 (42), which is important in regulating hepatic lipogenesis (43). This appears to be independent of insulin (42). Consistent with this, the expression of SREBP1c was reduced by 60% in liver of PTP1B Ϫ/Ϫ mice (Fig. 7K). SREBP1 expression in liver is higher in hyperinsulinemic and insulin-resistant states (43) and is increased by insulin signaling (43). Reduced insulin signaling in DHet mice tended to lower SREBP1c expression, although this did not reach statistical significance. The expression of liver SREBP1c was not different in DHet/PTP1B Ϫ/Ϫ and DHet mice (Fig. 7K).

DISCUSSION
We investigated the effects of reducing PTP1B expression on insulin sensitivity in a polygenic model of type 2 diabetes with similarities to human diabetes. In this model, diabetes results from a combination of relatively modest defects in insulin signaling (haploinsufficiency of IR and IRS-1) that act synergisti-  1 pY612; C), Akt/PKB Ser 473 (G), GSK3␤ Ser 21 /Ser 9 (H), and p70 S6K Thr 389 (I) in liver was measured by immunoblot analysis using antibodies specific for each phosphorylation site. In A-C, the upper bar graphs show total tyrosyl phosphorylation of IR and IRS-1 at specific sites, and the lower bar graphs show the specific tyrosyl phosphorylation corrected for IR and IRS-1 protein levels. Tyrosyl phosphorylation of IRS-2 was assessed by immunoprecipitation of liver lysates (500 g) with anti-IRS-2 antibody, followed by immunoblotting with anti-phosphotyrosine antibody (D). The immunoblots shown are representative of three blots for each panel. The bar graph shows quantitation of all three blots. IRS-1-associated (E) and IRS-2-associated (F) PI3K activities were measured as stated under "Materials and Methods." Phosphoenolpyruvate carboxykinase levels were measured using a rabbit polyclonal antibody against phosphoenolpyruvate carboxykinase (33) (J). Liver SREBP1c mRNA levels were measured by quantitative reverse transcription-PCR (K). For each genotype, n ϭ 3 for saline and n ϭ 6 -7 for insulin. ϩ, p Ͻ 0.05 versus saline (indicated for E and F only); *, p Ͻ 0.05 versus all other insulin-stimulated groups; #, p Ͻ 0.05 versus insulin-stimulated PTP1B Ϫ/Ϫ and DHet mice as determined by ANOVA.
cally to cause severe insulin resistance and diabetes (30). The severity of insulin resistance and the incidence of type 2 diabetes increase with age, which is also similar to type 2 diabetes in humans (30). Notably, the defects in this model parallel insulin signaling defects in humans with type 2 diabetes. For example, reduction of IR and IRS-1 levels to ϳ50% of those in normal subjects has been reported in muscle of obese insulin-resistant humans; this is associated with impaired insulin-stimulated IR and IRS-1 tyrosyl phosphorylation, IRS-1-associated PI3K activity, and glucose uptake (44). Decreased IR levels in adipocytes or circulating monocytes has also been reported in patients with impaired glucose tolerance and type 2 diabetes, and these defects are improved with antidiabetic therapy (45,46). The extent to which reversal of these signaling defects improves systemic insulin sensitivity is an important question. In this work, we report that deficiency of PTP1B improves insulin signaling in muscle and liver and reduces hyperinsulinemia, hyperglycemia, and the incidence of type 2 diabetes in a polygenic model of insulin resistance.
Unlike in other animal models in which PTP1B deficiency increases insulin sensitivity concurrent with a reduction in adiposity (21)(22)(23)(24), this study shows that, even without changes in body fat content, reduced expression of PTP1B can markedly improve insulin sensitivity and diminish the risk of developing type 2 diabetes. The absence of PTP1B in all tissues (23,24) or only in neurons (26) results in increased leptin sensitivity. Thus, it is possible that increased leptin sensitivity could contribute to improved insulin sensitivity in DHet/PTP1B Ϫ/Ϫ mice. Ambient leptin levels were lower in PTP1B Ϫ/Ϫ mice in this study, consistent with their enhanced leptin sensitivity shown previously (21,23,24). However, leptin levels in DHet/PTP1B Ϫ/Ϫ mice were not significantly lower than those in DHet and WT mice (supplemental Fig. 2). Because there is a range in insulin sensitivity in DHet/PTP1B Ϫ/Ϫ mice, we investigated whether mice with less hyperinsulinemia have higher leptin levels, but this was not the case (data not shown). Therefore, it is unlikely that enhanced leptin sensitivity in DHet/PTP1B Ϫ/Ϫ mice contributes to increased insulin sensitivity.
We also found that heterozygous deficiency of IR and IRS-1 results in differential effects on key steps in the insulin signaling cascade as well as on insulin signaling in muscle and liver (Figs. 6 and 7). PTP1B deficiency by itself results in enhanced insulinstimulated IR, IRS-1, and Akt/PKB phosphorylation and PI3K activity compared with WT levels. When PTP1B deficiency is superimposed on heterozygous deficiency of IR and IRS-1, insulin-stimulated IR phosphorylation is partially restored in muscle, consistent with increased insulin sensitivity at the level of IR in the setting of decreased IR protein levels. Even though IRS-1 protein is reduced by 60% in muscle of DHet/PTP1B Ϫ/Ϫ mice, the total amount of IRS-1 phosphorylated at Tyr 612 , the binding site for the p85 subunit of PI3K, is restored to WT levels, The stoichiometry of IR and IRS-1 phosphorylation after correction for the respective protein levels is also higher in muscle in PTP1B Ϫ/Ϫ mice compared with WT mice and in PTP1B-DHet mice compared with DHet mice, also consistent with increased insulin sensitivity. Increased phosphorylation of IRS-1 Tyr 612 is probably a critical factor in the improved insulin sensitivity in DHet/PTP1B Ϫ/Ϫ mice because activation of PI3K is a critical step for the metabolic effects of insulin in muscle and liver (41). Consistent with improved IRS phosphorylation, the impaired IRS-associated PI3K activity in muscle and liver of DHet mice is improved by PTP1B deficiency in DHet/ PTP1B Ϫ/Ϫ mice.
The total amount of insulin-stimulated IR phosphorylation is restored only partially in DHet/PTP1B Ϫ/Ϫ muscle compared with WT mice. Total IRS-1 phosphorylation and IRS-1-associated PI3K activity are improved to WT levels. Nevertheless, insulin-stimulated Akt/PKB phosphorylation in muscle of DHet/PTP1B Ϫ/Ϫ mice is increased above WT levels, to levels comparable with those in muscle of PTP1B Ϫ/Ϫ mice. A previous study suggests that Յ45% of maximal tyrosyl phosphorylation of IR and IRS-1 and Ͻ50% of maximal PI3K activity are sufficient to achieve maximal Akt/PKB phosphorylation and insulin-stimulated glucose uptake (47). Thus, less than complete normalization of IR may be sufficient to increase phosphorylation of downstream targets such as Akt/PKB above WT levels, especially if IR or IRS-1 phosphorylation is also sustained, as demonstrated by Elchebly et al. (21) for IR in liver of PTP1B Ϫ/Ϫ mice.
Alternatively, there could be novel PTP1B targets in the insulin signaling pathway distal to IRS-associated PI3K activity that modulate Akt/PKB activity. Insulin stimulation of PI3K activity generates phosphatidylinositol 3,4,5-trisphosphate, an important lipid second messenger required for activation of Akt/PKB (48). Cellular phosphatidylinositol 3,4,5-trisphosphate levels are also regulated by lipid phosphatases, including the 3Ј-lipid phosphatase PTEN (phosphatase and tensin homolog deleted on chromosome 10) and the 5-lipid phosphatase SHIP-2 (SH2 domain-containing inositol phosphatase 2), which dephosphorylate phosphatidylinositol 3,4,5-trisphosphate at the 3Ј-and 5Ј-positions, respectively (48). SHIP-2 can be tyrosyl-phosphorylated in response to interleukins and growth factors, including insulin (49,50). The regulatory effects of this phosphorylation are not known and may differ for different growth factor receptors and cell types (49,50). Potentially, the absence of PTP1B could decrease the activities of these lipid phosphatases, which could increase cellular phosphatidylinositol 3,4,5trisphosphate levels, leading to enhanced Akt/PKB phosphorylation. Other potential modifiers of Akt/PKB activity include TRB3, which has recently been shown to bind to Akt/PKB and to inhibit its phosphorylation by insulin (51).
Despite similar reductions of IR and IRS-1 levels in muscle and liver of DHet mice, IR and IRS phosphorylation in liver is less severely reduced than in muscle. However, Akt/PKB phosphorylation is reduced more in liver than in muscle. Previous data showed that the degree of impairment of insulin-stimulated Akt/PKB activation in liver of DHet mice depends on the genetic background, ranging from no impairment to an ϳ50% reduction (31). This reinforces the possibility that there can be important modulators of Akt/PKB activity downstream of IR and IRS-1. As evidence of downstream biologic effects of these alterations in insulin signaling, the expression of key genes involved in gluconeogenesis (phosphoenolpyruvate carboxykinase) and lipogenesis (SREBP1c) in liver is also altered. Phosphoenolpyruvate carboxykinase expression is up-regulated in liver of DHet mice, indicating hepatic insulin resistance and increased gluconeogenesis. This defect is corrected by PTP1B deficiency in DHet/PTP1B Ϫ/Ϫ mice. SREBP1c expression is reduced in liver of PTP1B Ϫ/Ϫ mice, as expected (42). It also tends to be lower in liver of DHet mice probably because of diminished insulin signaling resulting from heterozygous deficiency of IR and IRS-1. The lack of difference in SREBP1c expression in DHet/PTP1B Ϫ/Ϫ mice compared with DHet mice could be due to the opposing effects of PTP1B deficiency (independent of insulin) to decrease SREBP1c expression and improved insulin signaling to increase SREBP1c expression.
We also observed tissue-specific effects of IR/IRS-1 heterozygosity and PTP1B deficiency on other downstream signals. Whereas neither IR/IRS-1 heterozygosity nor PTP1B absence affects insulin-induced phosphorylation of GSK3 and p70 S6K in muscle, insulin-stimulated phosphorylation of both kinases is impaired in liver of DHet mice and is restored by PTP1B deficiency. These differential effects could reflect different upstream signaling pathways in these tissues. For example, IRS-2 plays a more significant role in IR signaling in liver, whereas IRS-1 is more important in muscle (30,52,53). Alternatively, the relative ability of other protein-tyrosine phosphatases to compensate for PTP1B deficiency or the differential action of other negative regulatory mechanisms (e.g. inhibitory serine/threonine phosphorylation, increased serine/threonine phosphatase activity) could also help to explain tissue-specific differences in the effects of PTP1B deficiency.
We also found that PTP1B deficiency reduces pancreatic islet enlargement. This most likely reflects the reduced peripheral insulin resistance, leading to diminished need for insulin secretion. Thus, our results demonstrate that reduction in PTP1B can markedly reduce the incidence of insulin resistance and frank diabetes in a polygenic model of insulin resistance and diabetes and also delay the onset of insulin resistance associated with aging. The therapeutic effects of PTP1B deficiency in this diabetic model strongly support the notion that inhibiting PTP1B activity will be effective even in people with a high genetic risk for type 2 diabetes. Furthermore, even if a pharmacologic inhibitor does not reduce adiposity (as would be the case for agents that do not access the central nervous system) (26), PTP1B inhibition should still be effective in enhancing insulin sensitivity and potentially preventing type 2 diabetes.