Liver-specific Protein-tyrosine Phosphatase 1B (PTP1B) Re-expression Alters Glucose Homeostasis of PTP1B–/–Mice*

Protein-tyrosine phosphatase 1B (PTP1B) is an important negative regulator of insulin and leptin signaling in vivo. Mice lacking PTP1B (PTP1B–/– mice) are hyper-responsive to insulin and leptin and resistant to diet-induced obesity. The tissue(s) that mediate these effects of global PTP1B deficiency remain controversial. We exploited the high degree of hepatotropism of adenoviruses to assess the role of PTP1B in the liver. Liver-specific re-expression of PTP1B in PTP1B–/– mice led to marked attenuation of their enhanced insulin sensitivity. This correlated with, and was probably caused by, decreased insulin-stimulated tyrosyl phosphorylation of the insulin receptor (IR) and IR substrate 2-associated phosphatidylinositide 3-kinase activity. Analysis using phospho-specific antibodies for the IR revealed preferential dephosphorylation of Tyr-1162/1163 compared with Tyr-972 by PTP1B in vivo. Our findings show that the liver is a major site of the peripheral action of PTP1B in regulating glucose homeostasis.

Insulin action is mediated by a complex network of signaling events (reviewed in Refs. 1 and 2). Upon binding to the insulin receptor (IR), 1 insulin induces autophosphorylation of several tyrosyl residues, leading to the recruitment and phosphorylation of insulin receptor substrates (IRSs), Gab family proteins, and the adapter Shc. These serve as docking sites for Src hyomology 2 domain-containing signal relay molecules, such as phosphatidylinositide 3-kinase (PI3K). PI3K is a major mediator of the metabolic actions of insulin, includ-ing its ability to promote glycogen synthesis and stimulate glucose transport (3).
Regulation of insulin action requires a balance between IR phosphorylation and dephosphorylation. Several PTPs have been implicated as negative regulators of insulin signal transduction. Chief among these is protein-tyrosine phosphatase 1B (PTP1B), a ubiquitously expressed, nonreceptor PTP localized on the endoplasmic reticulum (4 -6). Overexpression of PTP1B in various tissue culture cells decreases IR and IRS1 tyrosyl phosphorylation and reduces IRS1-associated PI3K activity (7)(8)(9). The IR and possibly IRS1 are direct PTP1B substrates (10 -14), and structural and kinetic studies suggest that PTP1B preferentially dephosphorylates the double phosphotyrosyl motif Tyr-1162/1163 in the IR (15). The physiological relevance of these observations was dramatically verified by mice lacking PTP1B (PTP1BϪ/Ϫ mice), which exhibit markedly increased insulin sensitivity and enhanced insulin signaling, with substantially increased IR and IRS1 tyrosyl phosphorylation in liver and muscle (16,17). Furthermore, hyperinsulinemic-euglycemic clamp studies reveal enhanced whole body glucose disposal in PTP1BϪ/Ϫ mice (17). Interestingly, this is due to elevated insulin-stimulated glucose uptake in skeletal muscle but not in white adipose tissue. PTP1BϪ/Ϫ mice also showed a trend toward increased insulin-evoked suppression of hepatic glucose production (17).
In addition to these effects on insulin signaling and glucose homeostasis, which are consistent with the earlier ex vivo studies, PTP1BϪ/Ϫ mice also display an unanticipated decrease in adiposity and resistance to high fat diet-induced obesity (16,17). Subsequent studies showed that PTP1B could regulate leptin signaling ex vivo, most likely by dephosphorylating Jak2, and PTP1BϪ/Ϫ mice showed increased hypothalamic leptin sensitivity (18,19). However, other studies indicate that, also by dephosphorylating Jak2, PTP1B negatively regulates hepatic growth hormone signaling, potentially providing a peripheral (non-central nervous system) explanation for decreased adiposity and resistance to diet-induced obesity (20). Furthermore, PTP1B antisense oligonucleotides, which lower PTP1B expression only in liver and fat, reportedly enhance insulin sensitivity in animal models of insulin resistance (21)(22)(23)(24). Thus, although there is general agreement that PTP1B is a major regulator of insulin sensitivity and body weight, the tissue(s) that mediate these effects have remained unclear.
We utilized the high degree of hepatotropism of adenoviruses to assess the effects of restoring PTP1B expression only in the livers of PTP1BϪ/Ϫ mice. PTP1BϪ/Ϫ mice expressing human PTP1B (hPTP1B) in the liver (at activity levels ϳ6-fold greater than in WT mice) exhibited dramatically decreased IR tyrosyl phosphorylation and IRS2-associated PI3K activity and reversal of their enhanced insulin sensitivity. We observed prefer-ential site-specific dephosphorylation of the IR at Tyr-1162/ 1163 by PTP1B in vivo, consistent with earlier in vitro studies (15). Our findings confirm that PTP1B is an important negative regulator of glucose homeostasis and strongly suggest that the liver is a major site of PTP1B action in the periphery.
Immunoprecipitation and Immunoblotting-The tissues were lysed in a modified radioimmune precipitation assay buffer (50 mM Tris-Cl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS and 5 mM EDTA), containing 2 mM sodium orthovanadate and a protease inhibitor mixture (final concentrations were 20 g/ml phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 1 g/ml aprotinin, 1 g/ml of pepstatin, and 1 g/ml of antipain). The lysates were clarified by centrifugation at 13,000 rpm for 10 min, and protein concentrations were determined using a bicinchoninic acid protein assay kit (Pierce). For immunoprecipitations, the lysates were incubated with the appropriate antibodies at 4°C for 3 h to overnight. Immune complexes were collected onto protein A-Sepharose beads, washed extensively, resolved by SDS-PAGE, and transferred onto Immobilon-P membranes (Millipore, Bedford, MA), which were blocked in 10 mM Tris-Cl, pH 7.4, 150 mM NaCl, 0.05% Tween 20 with either 5% bovine serum albumin or 5% Carnation nonfat dry milk. After incubation with appropriate primary and secondary antibodies (used at the concentrations recommended by their suppliers), the blots were visualized using ECL (Amersham Biosciences). Quantification was carried out by using National Institutes of Health Image Pro Software version 1.62.
Adenovirus Preparation and Animal Experiments-Adenoviruses encoding human PTP1B (AdPTP1B) and LacZ (AdLacZ) (8) were purified twice using the CsCl method and then titrated using cell viability assays, as described previously (25). The viruses were suspended in 200 l of phosphate-buffered saline and injected through the tail veins of 9 -12-week-old male mice at 3 ϫ 10 8 plaque-forming units/g of body weight. PTP1BϪ/Ϫ and wild type (WT) mice were weaned onto a low fat (chow) diet (product 5010; Purina, St. Louis, MO) and maintained on this diet for ϳ6 weeks. One week prior to virus injection, the mice were switched onto a high fat diet (product TD 93075; Harlan-Teklad, Madison, WI) and maintained on that diet until they were sacrificed. The low fat diet has 11.3% of calories from fat (physiological fuel value is 3.41 Kcal/g), whereas the high fat diet has 54% of calories from fat (physiological fuel value is 4.8 Kcal/g). Body weight and food intake were measured daily.
For insulin tolerance tests (ITTs), the mice were fasted for 4 h and then injected with 0.65 milliunits/g (body weight) human insulin (Novo Nordisk Pharmaceuticals, Princeton, NJ). Blood glucose values were measured immediately before and at 15, 30, 45, 60, and 90 min after insulin injection. To enable data comparison from different experiments, the values were expressed as percentages of change in blood glucose levels. For signaling experiments, the mice were fasted for 12-14 h and then injected intraperitoneally with 10 milliunits of insulin/g of body weight. The mice were sacrificed 10 min after injection, and insulin-responsive tissues were removed and snap frozen. Plasma insulin levels were determined by radioimmunoassay (Crystal Chem. Inc.; catalog number 90060). Plasma aspartate aminotransferase (AST) levels were measured at the Department of Laboratory Medicine (Children's Hospital, Boston, MA). Because the AST assay requires 100 l of serum, serial measurements were not possible. Thus, the AST levels were determined only at the end of the study. Adenoviral infection typically produces a mild hepatitis in mice. Accordingly, several injected animals had elevated AST levels. We excluded all infected mice (n ϭ 6 mice for KO AdPTP1B and n ϭ 1 mouse for WT AdLacZ) whose transaminases were more than three times elevated above the mean normal value (saline-injected WT mice). However, it should be noted that if all of the excluded mice were included, the results would not be changed, indicating that adenoviral infection of the liver itself did not alter glucose homeostasis. Generally high AST levels in KO AdPTP1B most likely reflect a role for PTP1B in mediating sensitivity to adenovirus-induced hepatitis. All of the experiments were approved by the Harvard Medical School Center for Animal Resources and Comparative Medicine and were conducted in accordance with the principles and procedures outlined in the National Institutes of Health Guide for Care and Use of Laboratory Animals.
Enzyme Activity Assays-PTP1B activity assays were carried out using reduced, carboxamidomethylated, and maleyated lysozyme phosphorylated with [␥-32 P]ATP, as described (26). The liver samples were lysed in 1% Triton X-100, 0.6 M KCl, 50 mM dithiothreitol, and the protease inhibitor mixture. Phosphatase assays were initiated by the addition of 10 l of radiolabeled carboxamidomethylated and maleyated lysozyme (10 M) and incubated at 37°C for 5 min. The reactions were terminated by the addition of 750 l of ice-cold acidic charcoal mixture (0.9 M NaCl, 90 mM sodium pyrophosphate, 2 mM NaH 2 PO 4 and 4% (v/v) Norit A). After centrifugation for 1 min, the amount of radioactivity in 400 l of supernatant was measured in a liquid scintillation counter. The data are represented as fold change compared with WT mice.
For PI3K assays, the tissue lysates (1.5 mg protein) were subjected to immunoprecipitation with polyclonal antibodies to IRS1 or IRS2, and immune complex PI3K activity was determined, as described previously (27).
Immunohistochemistry-Paraffin-embedded sections (5 m) were dewaxed twice in Xylene and then rehydrated through a series of 100, 95, 80, and 70% ethanol washes for 5 min each. The sections were boiled for 25 min in 1 mM EDTA for antigen retrieval, blocked with 2% normal donkey serum and 1% bovine serum albumin in phosphate-buffered saline, and then incubated in rabbit anti-human PTP1B antibodies (H-135) at 4°C overnight. Following washing, the sections were incubated with biotin-conjugated donkey anti-rabbit secondary antibodies (Jackson Immuno Research, West Grove, PA) for 1 h. For detection of bound antibodies, the sections were incubated in ABC complex (Vector Laboratories, Burlingame, CA) for 30 min, washed, and developed in 3,3Ј-diaminobenzidine for 5 min. The sections were then dehydrated, mounted, and dried overnight before observation.
Statistical Analysis-The data are expressed as the means Ϯ S.E. The statistical analyses were performed using the Statview program (Abacus Concepts, Berkeley, CA). ITTs were analyzed by repeated measures analysis of variance. Post hoc differences were considered significant at p Յ 0.05 and highly significant at p Յ 0.01 using a Fisher's protected least square difference. All of the other data were analyzed using analysis of variance.

Re-expression of PTP1B in PTP1BϪ/Ϫ Mice by Adenovirusmediated Gene
Transfer-To assess the role of PTP1B in the liver, we selectively re-expressed hPTP1B in the livers of PTP1BϪ/Ϫ mice by injecting them intravenously with recombinant Ad5 category viruses (28 -30). Male mice (9 -12 weeks old) on a low fat diet were switched to a high fat diet for 7 days, injected with the appropriate adenovirus, and then studied for an additional 7 days (Fig. 1A). hPTP1B protein expression in tissue lysates was assessed by immunoblotting with monoclonal antibodies against hPTP1B (FG6), which have low affinity for mouse PTP1B (31) (Fig. 1B). As expected, and consistent with the hepatotropism of these viruses, hPTP1B protein expression was present in the liver and absent from all other insulin-responsive tissues (25,28,29) (Fig. 1B). Immunohistochemical analysis (using H-135 anti-hPTP1B antibodies) revealed hPTP1B expression in a large number of hepatocytes, wherein it displayed the intracellular, reticular staining pattern characteristic of PTP1B (Fig. 1C, left panel) (4,6). Thus, hPTP1B is expressed in the expected location in infected hepatocytes. A similar staining pattern was observed using hPTP1B monoclonal antibodies (FG6) (data not shown), whereas no immunoreactivity was detected in the livers of AdLacZ-injected PTP1BϪ/Ϫ mice (Fig. 1C, right panel).
Because there are no antibodies available that recognize mouse and human PTP1B proteins equally, we could not easily compare the level of expressed hPTP1B to endogenous mouse PTP1B in WT mice by immunoblotting. Instead, we assessed PTP activity in liver lysates from WT AdLacZ, KO AdLacZ, and KO AdPTP1B mice. The overall PTP activity was ϳ20% less in liver of KO AdLacZ mice compared with WT mice. If we assume that all of the "missing" PTP activity is PTP1B, these data indicate that PTP1B normally accounts for ϳ20% of the total PTP activity in liver of WT mice. Upon re-expressing PTP1B in liver of the KO animals, the overall PTP activity was about 100% greater than that in WT mice. If all of this "extra" activity is due to (re-expressed) PTP1B, then the total PTP1B activity in the reconstituted mice is 120% of WT levels. In other words, the restored activity is 6-fold higher (120/20) than normal levels of activity.
Effects of Liver PTP1B Re-expression on Body Weight-We were interested in the potential effects of liver PTP1B on both glucose homeostasis and body mass. It is optimal to determine the effects of genetic manipulations on glucose homeostasis after puberty. Because PTP1BϪ/Ϫ mice have lower weight than WT littermates as early as 2 weeks post-weaning onto a high fat diet, by puberty such mice would weigh considerably less than WT controls. Such a weight difference would confound assessment of the effects of liver PTP1B on glucose homeostasis. Adenovirus-mediated gene expression in the liver is sustained for a relatively short time; thus, a long term study of the effect of hepatic PTP1B re-expression on weight gain is not feasible. Therefore to facilitate detection of possible differences in body weight caused by hepatic PTP1B expression, mice were placed on low fat diet after weaning, then switched to a high fat diet 1 week before virus administration, and maintained on that diet for the rest of the study (Fig. 1A). The mice were housed individually, and food intake and body weights were measured daily. Prior to viral injection (during the first 7 days post-switch to high fat diet), no significant differences in body weights of the mice were detected. Within a few days after viral injection, a significant difference was detected between KO AdLacZ and WT mice, but for most part there was no significant weight difference between mice in any of the virally injected groups ( Fig. 2A). As expected, there also was no significant difference in food intake between WT and PTP1BϪ/Ϫ mice prior to infection (17,18) (Fig. 2B). Viral infection had a mild but significant anorexogenic effect on both WT and PTP1BϪ/Ϫ mice. Interestingly, the duration of this suppressed food intake appeared to be longer in KO AdLacZ mice than in KO mice re-expressing PTP1B. However, all of these differences were small and had little effect on body mass and thus are unlikely alone to account for altered glucose homeostasis.
FIG. 1. Liver-specific expression of hPTP1B using adenovirus-mediated gene transfer. A, time line of the experiments. Mice (9 -12 weeks old) were switched to a high fat diet and then injected with adenovirus as described under "Materials and Methods. " ITTs, blood collections, and euthanasia were performed at the indicated times. B, specific reconstitution of hPTP1B expression in liver of PTP1BϪ/Ϫ mice. Lysates of liver, epididymal fat pad (white adipose tissue, WAT), brown adipose tissue (BAT), and skeletal muscle from PTP1BϪ/Ϫ mice injected with either AdPTP1B or AdLacZ and WT mice injected with AdLacZ were prepared as described under "Materials and Methods. " The lysates were subjected to SDS-PAGE and then immunoblotted for human PTP1B using monoclonal antibody FG6, followed by reprobing for SHP2 to ensure even loading. Lysates from two different PTP1BϪ/Ϫ mice re-expressing PTP1B in the liver are shown (the first three lanes are from one mouse, and the adjacent three lanes are lysates from another mouse). C, immunohistochemical localization of PTP1B in liver. The liver sections were prepared as described under "Materials and Methods" and stained for human PTP1B using polyclonal antihuman PTP1B antibodies, revealing PTP1B expression in the endoplasmic reticulum as expected. D, activity of reconstituted PTP1B in the liver. The liver lysates were prepared from KO AdPTP1B, KO AdLacZ, and WT AdLacZ mice. The lysates were subjected to phosphatase activity assay using 32 P-labeled carboxamidomethylated and maleyated lysozyme as a substrate. The data are presented as fold changes from WT where 100% corresponds to 1 pmol/min.

Re-expression of PTP1B in Liver of PTPBϪ/Ϫ Mice Attenu-
ates Insulin Sensitivity-PTP1BϪ/Ϫ mice have increased insulin sensitivity (16,17). Hyper-insulinemic euglycemic clamp studies revealed a marked enhancement of insulin-stimulated glucose uptake into skeletal muscle but also a trend toward greater suppression of hepatic glucose output in PTP1BϪ/Ϫ mice (17). Therefore, we assessed glucose homeostasis in PTP1BϪ/Ϫ mice after re-expression of hPTP1B in the liver. Consistent with previous studies, before viral injection, fed glucose levels were lower in PTP1BϪ/Ϫ (KO), compared with WT groups (Table I). There was no significant difference in fed or fasting glucose or fed insulin levels between PTP1BϪ/Ϫ mice and their counterparts re-expressing PTP1B (Table I). Serum AST levels were used as indications of liver function. PTP1BϪ/Ϫ mice infected with the PTP1B virus generally had higher AST levels than mice infected with the LacZ control virus. Moreover, PTP1BϪ/Ϫ mice infected with adenovirus LacZ had significantly lower AST elevations than WT mice injected with the same virus. When PTP1B expression is restored (adenovirus PTP1B into PTP1BϪ/Ϫ mice), AST levels were increased (compared with PTP1BϪ/Ϫ mice infected with adenovirus LacZ). The higher level of AST in PTP1BϪ/Ϫ mice infected with adenovirus PTP1B than in WT mice infected with adenovirus LacZ is likely explained by the 6 times higher level of PTP1B expression (compared with WT) in reconstituted mice. The data suggest that liver PTP1B levels may affect sensitivity to adenoviral induced hepatitis (see "Discussion").
To directly assess insulin tolerance in vivo, mice were subjected to ITTs. Fig. 3A shows the percentage of change in blood glucose values following insulin injection from two independent experiments containing mice from all of the indicated groups. These data were collected after viral infection. Notably, KO The arrow indicates the date of virus injection. B, food intake of individually housed mice was measured daily and presented as cumulative food intake (g) adjusted to body weight (g), before (left) and after (right) viral injection. The values depict the means Ϯ S.E. *, significant difference between KO AdPTP1B and KO AdLacZ; ϩ, significant difference between KO AdLacZ and WT; Ϫ, significant difference between KO AdPTP1B and WT; ϫ, significant difference between WT AdLacZ and WT. A single symbol (such as ϩ) indicates p Յ 0.05, whereas a duplicate symbol (such as ϩϩ) indicates p Յ 0.01.
AdPTP1B mice exhibited a significantly smaller decrease in blood glucose levels during ITT, compared with KO AdLacZ mice, indicating that insulin sensitivity is attenuated in PTP1BϪ/Ϫ mice re-expressing hPTP1B in the liver. Remarkably, the insulin sensitivity of KO AdPTP1B mice was similar to that of WT AdLacZ mice (Fig. 3A). The difference in the insulin response between KO AdLacZ and WT AdLacZ (and WT saline) was similar to the differences seen earlier in uninfected WT and PTP1BϪ/Ϫ mice (16,17), suggesting that the virus is not causing a major effect on insulin response.
To further test the effect of hepatic re-expression of PTP1B on insulin sensitivity, ITTs were carried out on the same cohort of mice before and after viral injection. Prior to injection and consistent with previous observations (16,17), PTP1BϪ/Ϫ mice were more insulin-responsive than WT mice (Fig. 3B). WT mice were injected with AdLacZ, whereas PTP1BϪ/Ϫ mice were injected with either AdLacZ or AdPTP1B, and then another ITT was performed (Fig. 1A). The insulin responses of KO AdLacZ and WT AdLacZ mice were not altered, compared with their initial response (Fig. 3C). In contrast, PTP1BϪ/Ϫ mice re-expressing PTP1B in liver were rendered significantly more insulin-resistant after viral delivery, and their insulin response was comparable with that of WT AdLacZ (Fig. 3C). Thus, hepatic selective re-expression of PTP1B is sufficient to alter the systemic insulin response, indicating that the liver is an important site of PTP1B action in the periphery.
Effect of PTP1B Re-expression on Insulin Signaling in Liver-To monitor the biochemical effects of liver-specific PTP1B expression, mice were injected with saline or insulin, and IR tyrosyl phosphorylation was assessed. Consistent with previous observations (16), the IR was hyperphosphorylated in KO AdLacZ compared with WT AdLacZ mice (Fig. 4A). Thus, adenoviral injection alone does not alter insulin-induced tyrosyl phosphorylation of the IR. In contrast, hepatic re-expression of hPTP1B in the liver of PTP1BϪ/Ϫ mice decreased IR tyrosyl phosphorylation (Fig. 4A).
IRS2 has a central role in mediating the actions of insulin in the liver, acting downstream of the IR and linking to PI3K activation (reviewed in Refs. 33 and 34). IRS1 plays a more minor role in the liver, although restoring IRS1 expression to the liver of IRS1Ϫ/Ϫ mice can correct insulin resistance in these mice (30). Insulin-stimulated IRS2-associated (Fig. 4C) and IRS1-associated (Fig. 4D) PI3K activity was increased 1.7-fold (p ϭ 0.04) and 2.2-fold (p ϭ 0.04) in KO AdLacZ mice compared with WT AdLacZ mice, respectively. Re-expressing hPTP1B led to a 3-fold (p ϭ 0.003) decrease in insulin-stimulated IRS2-associated PI3K activity compared with KO AdLacZ mice. The resulting levels of PI3K activity were nominally lower than that found in WT AdLacZ mice (Fig. 4C), but this did not reach statistical significance (p ϭ 0.24). IRS1-associated PI3K activity was decreased 2-fold in mice re-expressing hPTP1B compared with KO AdLacZ mice (Fig. 4D), but this also did not achieve statistical significance (p ϭ 0.07). Taken together, these data indicate that PTP1B is a major regulator of IR signaling and PI3K activity in the liver. DISCUSSION The primary sites of PTP1B action and the relative contribution of each site have remained controversial. PTP1BϪ/Ϫ mice exhibit decreased IR tyrosyl phosphorylation in liver and muscle, but not fat, suggesting that the former are potential sites for attenuation of IR signaling by PTP1B (16). The role of skeletal muscle is further supported by hyperinsulinemic-euglycemic clamp studies, which suggests that muscle is a major contributor to the increased insulin sensitivity of PTP1BϪ/Ϫ mice (17). However, several studies using PTP1B antisense oligonucleotides demonstrated that enhanced peripheral insulin sensitivity could be achieved by treatments that reduce PTP1B protein levels only in liver and fat but not muscle (21)(22)(23)(24). These oligonucleotides also caused significant weight reduction in ob/ob mice (21,22) to levels comparable with the decrease in body weight observed in compound ob/ob: PTP1BϪ/Ϫ mice (19).
Although these studies provided support for a role for PTP1B in regulating insulin signaling in the periphery, additional work was required to dissect its effects in various tissues. As a first step, we addressed the role of PTP1B in the liver. Our data indicate that liver-specific re-expression of hPTP1B in PTP1BϪ/Ϫ mice markedly attenuates their insulin hypersensitivity. Together with the finding that PTP1BϪ/Ϫ mice show increased tyrosyl phosphorylation of the IR and IRS-1 (Fig. 4, TABLE I Blood glucose, serum insulin, and AST levels in WT and PTP1BϪ/Ϫ mice infected with AdLacZ or AdPTP1B adenovirus Mice were fed a high fat diet ad libitum, and serum was collected from fed or overnight fasted mice. The indicated metabolic parameters were measured before and five days after viral delivery. The values are expressed as the means Ϯ S.E. Please note that for ease of comparison, the WT group is not included in the statistical analysis. A and B, and Ref. 16), increased IRS1-and IRS2-associated PI3K activity in liver (Fig. 4, C and D) and that these changes are reversed by liver-specific PTP1B expression, our results identify the liver as an important site wherein PTP1B acts to regulate glucose homeostasis. These data also are consistent with the effects of liver-specific IR knockout, which results in dramatic insulin resistance (35). Notably, Wang et al. (36) reported no change in whole body insulin responsiveness after liver-specific overexpression of PTP1B in rats. Although at first glance these results might appear to conflict with our findings, there are a number of plausible explanations. Expression of excess PTP1B in the liver of WT rats, where PTP1B already is highly expressed, might not be expected to alter insulin responsiveness; endogenous PTP1B levels may already be in excess of that necessary to regulate endogenous IR signaling in the liver. The results are expressed as the mean percentages of basal blood glucose concentration of mice from two independent experiments (9 -12 mice/group). B and C, assessment of insulin response in vivo by performing ITTs on the same group of mice before and after viral injection. B, ITTs on WT and PTP1BϪ/Ϫ mice before viral injection. The underlined, italic portion of the mouse genotype indicates the type of virus that would be subsequently injected. C, ITT was repeated on the same group of mice after viral injection. The values depict the means Ϯ S.E. *, significant difference between KO AdPTP1B and KO AdLacZ; , significant difference between KO AdLacZ and WT AdLacZ; ϩ, significant difference between KO AdLacZ and WT; Ϫ, significant difference between KO AdPTP1B and WT. In C, & indicates a significant difference between KO AdPTP1B before and after viral delivery. A single symbol (such as ϩ) indicates p Յ 0.05, whereas a duplicate symbol (such as ϩϩ) indicates p Յ 0.01. liver on a PTP1B-null background provides a more direct test of its normal role in this tissue. Alternatively, our studies were conducted while feeding mice a high fat diet, whereas Wang et al. used rat chow. Differences between PTP1B-expressing and nonexpressing rodents might be more apparent on this high fat diet background. Finally, we cannot exclude the possibility of interspecies differences in the role of hepatic PTP1B in regulating IR signaling and glucose homeostasis.
Insulin-stimulated IR tyrosyl phosphorylation is enhanced markedly in the liver of PTP1BϪ/Ϫ mice, and expressing hPTP1B reduces this hyperphosphorylation dramatically (Fig.  4A). We observed a higher level of IR hyperphosphorylation in our experiments than the 2-fold increase reported by Elchebly et al. (16), possibly because we measured IR phosphorylation at 10 min post-insulin injection rather than at the 5-min time point in their study. Alternatively, because the two PTP1BϪ/Ϫ lines are on different genetic backgrounds (129 ϫ C57Bl/6 versus 129 ϫ Balb), other loci may modify the effects of PTP1B deficiency.
Previous studies differ over whether PTP1B preferentially dephosphorylates specific sites on the IR. Seely et al. (11) found no difference in the ability of various IR-derived phosphotyrosyl peptides to compete with binding to the IR of a catalytically impaired "substrate-trapping" mutant of PTP1B. Based on these findings, they concluded that PTP1B had no site preference on the IR. Conversely, enzyme kinetic and structural analyses carried out by Salmeen et al. (15) indicated a marked preference of PTP1B for the bis-phosphotyrosyl motif at Tyr-1162/1163. Our data suggest that PTP1B can dephosphorylate both Tyr-1162/1163 and Tyr-972, and accordingly, both sites are hyperphosphorylated in the liver of PTP1BϪ/Ϫ mice. However, there is greater hyperphosphorylation of the bis-phosphotyrosyl motif compared with Tyr-972 in PTP1BϪ/Ϫ mice. These data provide in vivo support for the conclusions of FIG. 4. Attenuation of insulin receptor signaling in livers of PTP1B؊/؊ mice re-expressing PTP1B. A, tyrosyl phosphorylation of the IR in liver. The mice were injected intraperitoneally with saline or 10 milliunits of insulin/g of body weight and sacrificed 10 min after injection. The lysates were prepared as described under "Material and Methods," and IR immunoprecipitates (IP) were subjected to SDS-PAGE, immunoblotted for phosphotyrosine, and reprobed for IR. B, phosphorylation of the IR on Tyr-972 and Tyr-1162/1163. The liver lysates were subjected to SDS-PAGE and then immunoblotted with either IR Tyr-972 or Tyr-1162/ 1162 phosphospecific antibodies. The blots were scanned, and bands corresponding to phosphorylated IR were quantified using National Institutes of Health Image software. The results are the means of phosphorylated IR (from three independent experiments) normalized to levels of ERK. The data are presented as fold change compared with WT, where the value of 100% for Tyr-972 corresponds to 1.36 Ϯ 0.24 optical density units, and that for Tyr-1162/1163 corresponds to 1.48 Ϯ 0.35 optical density units. C and D, attenuation of IRS2-and IRS1-associated PI3K activity in PTP1BϪ/Ϫ mice re-expressing PTP1B in the liver. The liver lysates were subjected to immunoprecipitation with polyclonal antibodies to IRS1 or IRS2 and PI3K activity assay was performed as described. The values depict the means Ϯ S.E. *, significant difference between KO AdPTP1B and KO AdLacZ; #, significant difference between WT AdLacZ and KO AdPTP1B; , significant difference between KO AdLacZ and WT AdLacZ. A single symbol (such as ϩ) indicates p Յ 0.05, whereas a duplicate symbol (such as ϩϩ) indicates p Յ 0.01. Salmeen et al., although we cannot exclude alternative explanations for our findings. For example, PTP1B might dephosphorylate both Tyr-1162/1163 and Tyr-972 equally well, but the latter might be a better site than the former for another PTP. Our studies also show that increased IR tyrosyl phosphorylation results in increased IRS2-and IRS1-associated PI3K activity in PTP1BϪ/Ϫ compared with WT mice. Re-expression of PTP1B in the liver resulted in a decrease in IRS1-and IRS2-associated PI3K activity to levels indistinguishable from those of WT mice.
Re-expression of hPTP1B in the liver had little effect on body mass ( Fig. 2A). However, whether hepatic PTP1B has no effect on body mass regulation or rather that the duration of this study was too short to detect any such effect remains unclear. Interestingly, although there was no difference in food intake between any of the groups prior to viral infection, we observed an unexpected effect of hepatic PTP1B expression on food intake following viral infection. PTP1BϪ/Ϫ mice exhibited a greater anorexigenic effect after infection than WT mice, and hPTP1B expression appeared to abrogate this effect. Adenoviral infection typically produces a mild hepatitis, and indeed, serum transaminases were slightly elevated in our mice (Table I). KO AdPTP1B mice had higher serum transaminases than KO AdLacZ mice, so the differences in food intake do not appear to reflect the relative severity of hepatitis. Notably, Lam et al. (37) recently reported that a 5-fold hepatic overexpression of PTP1B in leptin-deficient mice resulted in severe impairment of the ability of leptin to reduce food intake; however, it did not alter body weight. Conceivably, leptin signaling in the liver somehow mediates the anorexigenic effect of adenoviral infection (38,39), although alternative explanations remain possible.
Another unexpected finding was that infecting PTP1BϪ/Ϫ mice with AdPTP1B consistently caused higher transaminase elevations than AdLacZ infection (Table I). This did not appear to reflect a greater sensitivity of PTP1BϪ/Ϫ mice to adenoviral infection, because KO AdLacZ mice had comparable serum transminases to WT AdLacZ mice. Instead, our data raise the intriguing possibility that PTP1BϪ/Ϫ mice may be relatively resistant to adenoviral-induced hepatitis. It has been reported that NFB mediates apoptosis in adenoviral hepatitis through the transcriptional activation of Fas (40). Regulation of Fas pathway by PTP1B is an interesting possibility, and further studies to assess the underlying mechanism seem warranted.
Whereas our data indicate that the liver is a substantial contributor to the insulin hypersensitivity of PTP1BϪ/Ϫ mice, it is unlikely to be the only tissue involved. Several recent studies utilizing tissue-specific gene disruption have revealed intercommunication between various insulin target tissues (41). For example, intracerebroventricular administration of leptin improves hepatic insulin sensitivity, probably through an efferent pathway from the central nervous system (42,43). Overexpressing PTP1B in the muscle of WT mice leads to insulin resistance in muscle but also decreases insulin-stimulated PI3K activity in the liver (31). This indirect effect of PTP1B expression suggests muscle-liver communication via secreted factors or neural transmission. The fact that hPTP1B is expressed at higher than normal levels in our reconstitution studies also limits the extent to which quantitative conclusions can be drawn about the relative contribution of PTP1B in different tissues to glucose homeostasis and body mass regulation. Additional experiments, such as tissue-specific deletion of PTP1B, will be required to precisely delineate these roles. In this regard, preliminary studies of mice with selective PTP1B deletion in liver show no effect on body weight but effects on glucose homeostasis consistent with the data presented here. 2 Several pharmaceutical companies are attempting to develop small molecule inhibitors of PTP1B for the treatment of insulin resistance/ type II diabetes. Our results suggest that agents that target PTP1B only in the periphery may have significant efficacy.