The pseudophosphatase phogrin enables glucose-stimulated insulin signaling in pancreatic β cells

Autocrine insulin signaling is critical for pancreatic β-cell growth and activity and is at least partially controlled by protein-tyrosine phosphatases (PTPs) that act on insulin receptors (IRs). The receptor-type PTP phogrin primarily localizes on insulin secretory granules in pancreatic β cells. We recently reported that phogrin knockdown decreases the protein levels of insulin receptor substrate 2 (IRS2), whereas high-glucose stimulation promotes formation of a phogrin–IR complex that stabilizes IRS2. However, the underlying molecular mechanisms by which phogrin affects IRS2 levels are unclear. Here, we found that relative to wildtype mice, IRS2 levels in phogrin-knockout mice islets decreased by 44%. When phogrin was silenced by shRNA in pancreatic β-cell lines, glucose-induced insulin signaling led to proteasomal degradation of IRS2 via a negative feedback mechanism. Phogrin overexpression in a murine hepatocyte cell line consistently prevented chronic insulin treatment–induced IRS2 degradation. In vitro, phogrin directly bound the IR without the assistance of other proteins and protected recombinant PTP1B from oxidation to potentiate its activity toward the IR. Furthermore, phogrin expression suppressed insulin-induced local generation of hydrogen peroxide and subsequent PTP1B oxidation, which allowed progression of IR dephosphorylation. Together, these results suggest that a transient interaction of phogrin with the IR enables glucose-stimulated autocrine insulin signaling through the regulation of PTP1B activity, which is essential for suppressing feedback-mediated IRS2 degradation in pancreatic β cells.

Pancreatic ␤ cells synthesize and secrete the pivotal hormone insulin, which regulates glucose homeostasis and nutrient storage by activating its specific receptor in central and peripheral tissues. ␤ cells express insulin receptor (IR) 3 and its downstream signal pathway components, and genetic studies showed that several elements of this pathway play an essential role in ␤-cell growth and function (1)(2)(3). Mice lacking IR in ␤ cells exhibit decreased levels of glucose-stimulated insulin secretion (GSIS) as well as age-dependent reductions in ␤-cell mass, whereas insulin receptor substrate 2 (IRS2) knockout mice develop diabetes due to a defect in compensatory ␤-cell proliferation (4 -6). Importantly, expression of insulin signaling proteins is reportedly decreased in islets from patients with type 2 diabetes (7).
Despite the growing evidence of the significance of insulin signaling in ␤ cells, whether secreted insulin acts on its own receptor in an autocrine/paracrine fashion remains controversial (8). Insulin signaling in response to glucose was impaired in ␤-cell lines derived from IR-or IRS2-knockout mice (9) and in cells of the mouse insulinoma line MIN6 that have siRNA-induced silencing of IR expression (10). Additionally, antibodies specific to insulin could abrogate glucose-stimulated MIN6 proliferation (11), which suggests that secreted insulin has an autocrine or paracrine effect on pancreatic ␤ cells (12,13). On the other hand, exogenous insulin rather than glucose stimulation reportedly promotes proliferation of primary mouse islet ␤ cells (14), and only low concentrations of insulin could protect primary islet cells from apoptosis (15). However, there is the possibility that ␤ cells could be exposed to high levels of insulin or continuously stimulated by extracellular insulin within the islet. If this is the case, chronic insulin stimulation or high local concentrations of insulin may drive ␤ cells into an insulin-resistant state through desensitization of the insulin-signaling pathway (8).
In the IR-and IRS-signaling pathways, protein-tyrosine phosphatases (PTPs), such as PTP1B and TCPTP, are largely responsible for returning cells to a basal state by dephosphorylating tyrosine-phosphorylated IRs or IRSs (16,17). In pancreatic ␤ cells, PTP1B is reportedly involved in dephosphorylating IR (18). Like other tyrosine receptors, IR activation induces reactive oxygen species (ROS) generation, presumably by inactivating cytosolic peroxiredoxins and activating membrane-anchored NADPH oxidases, and the resulting ROS can specifically and transiently inhibit PTP activity by reversible oxidation (19,20). Because IR phosphotyrosines are dephosphorylated by restored PTPs, PTP oxidation probably plays an important role in determining the duration of insulin signaling. However, prolonged insulin stimulation can overcome PTP regulation and result in serine/threonine phosphorylation of IRSs that is mediated by downstream kinases, including Akt, mTOR, and PKC (21). Consequently, IRSs are ubiquitinated and then degraded by the proteasome, thereby preventing further insulin signaling. The desensitization by this negative regulation seems to reflect an insulin-resistant state. Previous reports demonstrated that chronic exposure (Ͼ8 h) to high glucose in pancreatic ␤ cells induces IRS2 degradation by a negative feedback pathway (22,23), as was shown in adipocytes or hepatocytes.
Recently, we proposed one receptor-like PTP as a regulator of autocrine insulin signaling in pancreatic ␤ cells (24). This protein, phogrin (also known as PTPRN2 or IA-2␤), is enriched in insulin granules, where it contributes to glucose-stimulated ␤-cell growth by stabilizing IRS2 expression. Phogrin translocates to the plasma membrane and interacts with IR when insulin exocytosis is induced by glucose. This phogrin-IR interaction is probably functional, because cell growth retardation and degradation of IRS2 by phogrin knockdown were not observed in IR-deficient cells (24). Mice having a global gene deletion of phogrin exhibit mild glucose intolerance and a slight decrease in islet insulin content (25,26) but no alterations in ␤-cell proliferation. On the other hand, the phogrin homolog IA-2 (also known as PTPRN or ICA512) protein also presents a growthpromoting function in both ␤-cell lines and mouse islets (24,27). Specific knockdown of IA-2 in MIN6 and INS-1E ␤-cell lines reduces the cell proliferation rate, whereas IA-2-deficient mice have low ␤-cell regeneration rates after partial pancreatectomy. Phogrin and IA-2 show the highest sequence similarity in the cytoplasmic PTP domain and lack phosphatase activity for general PTP substrates because of amino acid mutations in the evolutionarily conserved catalytic domain (28). In contrast, the N-terminal pro-domain and the membrane-proximal mature domain called matN (also called the ectodomain) are involved in targeting to secretory granules in endocrine cells (29,30).
We previously showed that phogrin knockdown decreased IRS2 protein levels and promoted growth retardation in MIN6 cells and cultured mouse islets. We also showed that phogrin formed a complex with activated IR, although whether phogrin affects the tyrosine phosphorylation status of IR is unclear. Here, we extend our studies to explore how phogrin regulates insulin signal transduction for pancreatic ␤-cell growth. We show that phogrin supports PTP1B activity to achieve an overall regulation of glucose-stimulated insulin signaling in ␤ cells.

IRS2 protein in islets is down-regulated by phogrin deletion
To assess the effect of phogrin deletion on pancreatic ␤ cells, phogrin flox/flox mice (Fig. 1A) were crossed with rat insulin promoter (RIP)-cre mice to produce conditional knockout mice that have phogrin expression depletion in ␤ cells. Immunohistochemical analysis with phogrin-specific antibodies showed marked reduction of phogrin expression in the center region of islets from RIP-cre ϩ/Ϫ _phogrin flox/flox (knockout (KO)) mice relative to islets from control (Cre ϩ/Ϫ _phogrin ϩ/ϩ ) mice (Fig.  1B). Although pancreatic ␣ cells were likely to be stained by the phogrin antibody, nearly complete depletion of phogrin protein in islets of KO mice was observed by immunoblotting analysis (Fig. 1C). In contrast, only a moderate decline in phogrin expression was observed in the hypothalamus. The body weights, blood glucose concentrations, and plasma insulin levels were indistinguishable between male control and KO mice (data not shown).
The morphology of pancreatic islets was unchanged by phogrin knockout as analyzed by hematoxylin-eosin staining (not shown), and the ␤-cell mass per pancreas was similar between 16-week-old control and KO mice as assessed by immunostaining with insulin antibody (0.503% versus 0.493%). Although phogrin may not affect development of islet ␤ cells in mice, the incorporation rate of [ 3 H]thymidine in KO islets was slightly less than that of control islets (Fig. 1D). Meanwhile, low glucose stress-induced apoptosis was increased in primary ␤ cells prepared from KO islets compared with those from control islets ( Fig. 1E and Fig. S1). Importantly, adenovirus-mediated expression of phogrin completely restored apoptosis levels to that of control cells. We next examined expression levels of phogrin-associated proteins in the islets of control and KO mice. IRS2 levels in KO mouse islets were consistently lower than those of control mice at different ages ( Fig. 1F and Fig. S2). This result suggests that the proliferative activity of pancreatic ␤ cells is decreased by phogrin knockout via down-regulation of IRS2 protein levels. A slight reduction in IA-2 protein expression was similarly observed in phogrin-deficient islets, but there were no significant changes in other insulin granule proteins, such as carboxypeptidase E (CPE), secretogranin III (SgIII), Rab27, and VAMP2 (Fig. 1F).

Phogrin functions in IRS2-mediated insulin signaling
Silencing of phogrin in pancreatic ␤ cells reportedly induces ubiquitin proteasome-mediated degradation of the IRS2 protein (24). To elucidate the molecular basis of IRS2 down-regulation, we examined whether phogrin modulates glucose-dependent IRS2 increases at mRNA and protein levels using rat insulinoma INS-1E cells. Consistent with results of previous studies (31), high glucose (HG; 25 mM) promoted a significant increase in both IRS2 mRNA and protein levels in control cells (expressing shVector) ( Fig. 2A). In contrast, IRS2 protein did not increase in response to HG stimulation in phogrin-silenced cells (shPh3), whereas elevation of mRNA levels was unaffected by phogrin knockdown (Fig. 2A). Because chronic exposure (Ͼ8 h) of pancreatic ␤ cells to HG induces proteasomal degradation of IRS2 through a negative feedback mechanism (phos-

Phogrin connects insulin secretion and signaling
phatidylinositol 3-kinase (PI3K)-, Akt-, and mTOR-mediated serine phosphorylation of IRS2 and subsequent ubiquitination) (22), we examined whether IRS2 protein levels are reduced by the predicted feedback regulation in the phogrin-knockdown cells. Again, HG culture (4 h)-induced IRS2 protein accumulation was not observed in the shPh3-infected mouse ␤-cell line MIN6, and in phogrin-silenced cells, IRS2 levels in the HG culture were comparable with that of cells incubated in low glucose (LG; 2 mM) (Fig. 2B). IRS2 accumulation in phogrin-silenced cells was completely restored by treatment with the PI3K inhibitors wortmannin and LY294002, the mTOR inhibitor rapamycin, or the c-Jun N-terminal kinase (JNK) inhibitor SP600125 (Fig. 2B, right graph). Therefore, decreased IRS2 protein in the phogrin-knockdown cells may have resulted from proteasomal degradation via PI3K/mTOR/JNK-mediated signals. Together, these findings suggest that phogrin functions in glucose-regulated insulin-signaling pathways but does not directly stabilize IRS2 protein in pancreatic ␤ cells.
Ubiquitin proteasome-mediated degradation of IRS2 in negative feedback regulation was observed in various insulin-sensitive cells (32). To confirm the involvement of phogrin in insulin signaling, we investigated the effects of phogrin expression in a SV40-transformed mouse hepatocyte cell line (mHEPA). Because non-endocrine cells such as hepatocytes do not express prohormone-processing enzymes and lack hormonecontaining secretory granules, for these assays, we used an adenovirus expressing the pro-domain-deleted mature form of phogrin (⌬Pro-phogrin). In control cells, prolonged treatment (ϳ8 h) with insulin caused a 45% reduction in levels of the IRS2 protein, whereas chronic insulin treatment did not induce decreases in IRS2 protein in mature phogrin-overexpressing mHEPA cells (Fig. 2C). This result demonstrated the potential of phogrin to protect IRS2 from insulin-induced degradation not only in pancreatic beta cells but also in other insulin-sensitive cells.

Phogrin binds to phosphorylated insulin receptor
We previously showed that IRS2 protein expression levels correlated with molecular interactions between phogrin and IR, which are induced by the translocation of phogrin to the plasma membrane of ␤ cells exposed to HG conditions (24). To confirm that transient interactions occur between phogrin and IR in living cells, we used an imaging system based on reconstitution of two split fragments of mKG fluorescent protein (33). Fluorescent signals for mKG were observed only when IR-KGN and phogrin-KGC were cotransfected into MIN6 cells (Fig. S3). We then examined the level of phogrin/IR interaction by a co-immunoprecipitation assay using MIN6 cell lysates. Both the IR kinase inhibitor HNMPA(AM) 3 and the PTP inhibitor vanadate blocked the phogrin/IR interaction (data not shown), suggesting that the molecular interaction between IR and phogrin requires both autophosphorylated tyrosine in the IR ␤ subunit and the PTP structure in the phogrin cytoplasmic tail. To determine the phogrin-binding site of IR, three mutants were constructed. In each mutant, the indicated tyrosine(s) was substituted with non-phosphorylatable phenylalanine: Y1F, Tyr-960 of the juxtamembrane domain; Y3F, Tyr-1146, Tyr-1150, and Tyr-1151 in the catalytic domain; Y2F, Tyr-1316 and Tyr-1322 in the C-terminal region. Each IR mutant was EGFP-tagged and expressed in MIN6 cells for coimmunoprecipitation assays. Although the Y2F mutant did not overexpress well (Fig. 3A, left), the protein content of Y2F and Y3F but not Y1F in the phogrin immunoprecipitates was significantly decreased compared with that of IR-EGFP WT (Fig. 3A, right). This result suggests that phosphorylated tyrosines in Y2 and Y3 sites are involved in phogrin binding.

Phogrin enhances the ability of PTP1B to dephosphorylate the insulin receptor
Because phogrin and IA-2 both possess an inactive PTP domain in their cytoplasmic regions, we investigated the effects of phogrin on the IR phosphorylation status. We first analyzed whether recombinant phogrin interacts with tyrosinephosphorylated IR in vitro. Cell extracts from insulin-treated COS7 cells expressing IR-EGFP were adsorbed with anti-GFPagarose. The precipitated IR-EGFP was then incubated with purified glutathione S-transferase (GST)-fused PTP proteins, and bound proteins were analyzed by immunoblotting with anti-GST antibody. Both phogrin and IA-2 specifically bound to tyrosine-phosphorylated IR-EGFP, and not to EGFP (data not shown), but this binding was less efficient than TCPTP (Fig. 3B). The region within phogrin required for binding to IR was mapped to the cytoplasmic PTP portion (CT). We simultaneously analyzed the tyrosine phosphorylation levels of IR-EGFP with an anti-phosphotyrosine antibody. TCPTP, but neither phogrin nor IA-2, showed dephosphorylation activity against tyrosine-phosphorylated IR-EGFP, although all three proteins were found to interact with this substrate (Fig. 3C). We next examined PTP enzyme activity using p-nitrophenyl phosphate (pNPP) as a substrate. Again, phogrin and IA-2 showed no phosphatase activity in vitro (Fig. 3D), which is consistent with previous reports (34). Unexpectedly, however, the mature forms of phogrin and IA-2, but not their truncated versions (matN and CT), enhanced PTP activity of PTP1B by 1.6-and Phogrin connects insulin secretion and signaling 1.5-fold, respectively (Fig. 3D). An assay using phosphorylated IR-EGFP also showed this enhancing activity of phogrin (Fig.  3E), which indicated that phogrin facilitates dephosphorylation of IR only in the presence of PTP1B. In contrast, phogrin did not affect IR tyrosine kinase activity in an in vitro IR autophosphorylation assay (data not shown). The effect of phogrin on IR tyrosine phosphorylation was next explored using ␤ cells and non-␤ cells. First, we assessed phogrin overexpression using an mHEPA hepatocyte cell line. Insulin treatment of mHEPA cells promptly led to tyrosine phosphorylation of IR, and IR dephosphorylation began after a 10-min incubation in LacZ-expressing control cells (Fig. 4A, left panels). Exogenous expression of mature phogrin hastened IR dephosphorylation, and the phosphorylation content of IR decreased more rapidly (Fig. 4A, right panels). Rat insulinoma INS-1E cells were next used to examine autocrine insulin signaling. HG stimulation in INS-1E cells induced tyrosine phosphorylation of IR by 5-10 min, and thereafter, dephosphorylation occurred and returned to basal levels by 20 -30 min (Fig.  4B). In phogrin-overexpressing cells, the phosphorylation content of IR decreased more rapidly, returning to basal levels within 20 min. Phosphatases do not simply act to terminate a signal but practically contribute to setting the signaling response (35). Thus, our results suggest that phogrin contributes to autocrine insulin signaling by regulating the duration of IR phosphorylation in pancreatic ␤ cells.

Phogrin protects PTP1B from oxidation
Catalytic cysteine residues in PTPs are susceptible to ROS generated upon cell-surface receptor activation, and reversible oxidation of PTPs is involved in downstream signaling (16,20). We thus examined whether phogrin affects oxidation of recombinant PTP1B. In vitro, H 2 O 2 treatment (0.1-1 mM) promoted formation of recombinant PTP1B aggregates, which were undetectable by immunoblotting with anti-PTP1B under nonreducing conditions (Fig. 5A, top left). However, in the presence of reducing agent (2-mercaptoethanol), immunoblotted signals were restored to the control levels ( Fig. 5A, top right). When GST-phogrin was added to the assay, PTP1B oxidation was significantly attenuated even in the presence of 1 mM H 2 O 2 (Fig.  5A, left panel and graph). We next investigated intracellular ROS levels by incubating pancreatic ␤ cells with the oxidant probe 5-(and-6-)chloromethyl-2Ј,7Ј-dichlorodihydrofluorescein diacetate (CM-H 2 DCFDA). Phogrin knockdown in MIN6 cells enhanced intracellular ROS levels, as indicated by green fluorescence of DCF (Fig. 5B). Consistently, only weak fluores-cence signals were observed in mature phogrin-overexpressing (⌬Pro-phogrin) mHEPA hepatocyte cells, although insulin treatment evoked rapid H 2 O 2 production in control (LacZ) cells (Fig. 5C). These data suggest that phogrin prevents insulin-stimulated H 2 O 2 production. We further examined the effect of phogrin expression on insulin-induced oxidation of intracellular PTP1B. Mature phogrin-overexpressing and control mHEPA cells were treated with insulin, and endogenous PTP1B was then immunoprecipitated from the cell lysates. Immunoblotting with anti-Oxi-PTP antibody revealed that phogrin can indeed prevent insulin-induced oxidation of PTP1B in hepatocytes (Fig. 5D).

Discussion
Insulin resistance in pancreatic ␤ cells is frequently associated with a decrease in ␤-cell mass that contributes to type 2 diabetes pathophysiology (3,36). The finding that chronic stimulation with a high amount of insulin induces proteasomal degradation of IRS2 protein by a negative feedback pathway (22) suggested that insulin released in response to glucose stimulation of ␤ cells does not promote proliferation by the autocrine action of insulin (8). This condition is considered to be an insulin-resistant state in most insulin-sensitive cells. Meanwhile, genetic studies with mice provide evidence that the insulinsignaling pathway plays a critical role in ␤-cell growth and function (4 -6, 37). Furthermore, a compensatory proliferation of ␤ cells in mice is largely dependent on glycolysis and membrane depolarization (38), suggesting that GSIS is coupled to glucosestimulated ␤-cell growth. Therefore, if autocrine insulin actions are presumed to occur, ␤ cells might possess specific proteins or mechanisms that suppress the negative feedback pathway of insulin signaling. Because phogrin, an autoantigen of insulin-dependent diabetes mellitus, is highly expressed in pancreatic ␤ cells but not in hepatocytes, adipocytes, and myocytes (28), it may act as a specific regulator of autocrine insulin signaling in pancreatic ␤ cells.
We previously showed that phogrin forms a complex with IR in MIN6 cells when it translocates to the plasma membrane during GSIS activation (24). The interaction of phogrin with IR leads to stabilization of the IRS2 protein that can promote ␤-cell growth; however, the specific mechanisms of this function are undefined. Here we demonstrate for the first time that phogrin directly binds to IR without the assistance of any other molecules ( Fig. 3 and data not shown). Mutation analyses of IR phosphotyrosine residues and co-immunoprecipitation experiments with pharmacological inhibitors suggested that the PTP

Phogrin connects insulin secretion and signaling
structure of phogrin is primarily involved in its binding to the tyrosine-phosphorylated IR ␤-subunit ( Fig. 3A and data not shown). A previous structural study of PTP members demon-strated that the secondary substrate-binding site of the NT1 subgroup represented by PTP1B and TCPTP is distinct from that of the R8 IA-2 family subgroup (39). Indeed, PTP1B targets

Figure 4. Phogrin accelerates IR dephosphorylation in both hepatocytes and pancreatic ␤ cells.
A, mHEPA hepatocyte cells were infected with LacZ or ⌬Pro-phogrin adenoviruses. After culturing in serum-free MEM␣ containing 0.5% BSA for 12 h, cells were stimulated with 100 nM insulin for the indicated time. Tyrosine phosphorylation state of IR in cell extracts was determined by immunoblotting of anti-phosphotyrosine (pY20) immunoprecipitates (IP) with anti-IR␤ antibodies. The intensity of each band was quantified by densitometry, and the results are presented as -fold increase Ϯ S.E. (error bars) (n ϭ 3) relative to the control (time 0) (bottom graph). IR and phogrin expression levels were determined by immunoblotting. B, INS-1E cells infected by LacZ or phogrin adenoviruses were cultured in serum-free RPMI containing 2 mM glucose and 0.5% BSA for 12 h, preincubated in LG for 2 h, and then stimulated with HG for the indicated time. IR was immunoprecipitated from cell extracts and assessed by immunoblotting with anti-phosphotyrosine antibody (pY20). The same membrane was reprobed with anti-IR␤ antibody for a loading control. The intensity of each band (pY20) was quantified by densitometry, and the results are presented as -fold increase Ϯ S.E. (n ϭ 4) relative to the control (time 0) (bottom graph).

Phogrin connects insulin secretion and signaling
the phosphotyrosine in the juxtamembrane Y1 site of IR ␤-subunit for dephosphorylation (40), whereas mutation of this tyrosine residue did not affect phogrin-IR binding (Fig. 3A). This result may explain why phogrin does not compete with PTP1B (TCPTP) for binding to IR (data not shown). Direct interaction of phogrin with IR neither competes with nor inhibits PTP1Bmediated IR dephosphorylation, but phogrin instead potentiates PTP1B phosphatase activity (Fig. 3, D and E). Notably, the enhancement of PTP1B activity by phogrin required the mature form of phogrin that includes the matN extracellular region, the transmembrane domain, and the cytoplasmic PTP region (CT), whereas the CT fragment alone was sufficient for IR binding (Fig. 3, B and D). Our coimmunoprecipitation and in vitro assays confirmed that phogrin does not directly bind PTP1B (data not shown). These results indicate that molecular interactions of phogrin with IR on the plasma membrane could contribute to spatiotemporal interactions between phogrin and PTP1B in pancreatic ␤ cells. As such, phogrin probably contributes to the enzymatic activity of PTP1B by protecting it from ROS-induced oxidation (Figs. 3 (D and E) and 5 (A  and D)). However, because we did not identify the oxidative form of phogrin in our assays, the molecular mechanisms by which phogrin performs antioxidant functions require further investigation.
The phogrin-dependent suppression of ROS generation could potentiate PTP1B phosphatase activity, which results in

Phogrin connects insulin secretion and signaling
prompt IR dephosphorylation (Fig. 4). This transient activation of insulin signaling forms the basis for autocrine insulin action in pancreatic ␤ cells (Fig. 6B). When ␤ cells lack phogrin, very high levels of ROS that are probably induced by secreted insulin inactivate PTP1B through irreversible oxidation of the catalytic cysteine. Consequently, autocrine insulin signaling induces a persistent activation of downstream signals (Fig. 6A, heavy line) that in turn leads to feedback IRS2 down-regulation, as was previously predicted (8). Our earlier data indicated that shRNA-mediated phogrin reduction in ␤-cell lines or isolated islets led to decreased IRS2 protein levels (24), and here we found that phogrin silencing-induced decreases in IRS2 depend on HG stimulation as well as PI3K-, mTOR-, and JNKmediated negative feedback mechanisms (Fig. 2B). Furthermore, we observed a 44% reduction in IRS2 expression in islets derived from phogrin-knockout mice (Fig. 1F). Taken together, our results suggest that phogrin plays an essential role in protecting IRS2 protein from degradation, which allows HG-induced autocrine insulin signaling to proceed (Fig. 6B). Meanwhile, when ␤ cells are exposed to LG conditions, phogrin is sequestered in secretory granules, and PTP1B protection does not occur. Therefore, only low concentrations of exogenous insulin promote islet survival in culture, whereas high concentrations of insulin presumably induce oxidative inhibition of PTP1B and the associated feedback down-regulation of IRS2 (15).
In contrast to phogrin, STAT/cyclin-mediated mechanisms are involved in the growth-promoting function of IA-2 (ICA512) (27), suggesting that these two homologous proteins have different functions. Indeed, IA-2 expression in phogrindeficient islets was slightly decreased rather than increased (Fig. 1F). This line of thinking is supported by previous studies showing that ␤-cell proliferation after partial pancreatectomy is reduced in IA-2-deficient mice (27), whereas IRS2 appears not to have a role in this proliferation (41). Thus, at a mechanistic level, phogrin and IA-2 differ, although their overall regulatory effects on ␤-cell growth are similar.
Mice that are specifically deficient in IRS2 in ␤ cells and the hypothalamus exhibit an age-dependent reduction in the ␤-cell mass (42,43), whereas IRS2 is crucially involved in ␤-cell proliferation in response to high-fat diet-induced insulin resistance (44). A possible explanation of the unchanged ␤-cell mass in phogrin-knockout mice is that phogrin regulates ␤-cell growth only as a compensatory response. Our preliminary experiments show that phogrin-deficient mice fed the highfat diet had significantly lower compensatory ␤-cell hyperplasia than did control mice. 4 We presume that phogrin stabilizes the IRS2 protein through interactions with the IR only in response to demands of ␤-cell mass expansion. Thus, additional studies are needed to elucidate the role of phogrin in ␤-cell growth.
The present data support our hypothesis that phogrin is a unique regulator of the autocrine insulin signal pathway in pancreatic ␤ cells. Moreover, our data support the existence of a direct autocrine action of insulin, although several questions remain, including what is the behavior of secreted insulin around ␤ cells. We found that microlocalization of phogrin-EGFP at the cell surface overlapped with the punctate signals produced by staining with an anti-IR␣ subunit antibody (data not shown). Further investigation using recently developed super-resolution microscopy techniques could help visualize extracellular insulin behavior in greater detail, which would provide a further characterization of IR activation microdomains and exocytotic domains of insulin granules.

Antibodies and reagents
The anti-phogrin and IA-2 rabbit polyclonal antibodies against the luminal region were affinity-purified (24). These antibodies were previously characterized (24,45). The guinea pig anti-insulin antibody, anti-␤-actin, and anti-GST mouse monoclonal antibodies were purchased from Sigma. Antiphosphotyrosine mouse (pY20), anti-CPE, and anti-pan-ERK mouse monoclonal antibodies and anti-IR␤ rabbit polyclonal antibodies were purchased from BD Biosciences (Lexington, KY). Anti-oxidized PTP active site (Oxi-PTP) mouse monoclonal antibody and anti-PTP1B rabbit polyclonal antibodies were from R&D Systems (Minneapolis, MN). Anti-IRS2 rabbit, anti-Rab27 rabbit, and anti-GFP mouse monoclonal antibodies were from Upstate Biotechnology, Inc. (Lake Placid, NY), Immuno- A, in insulin-sensitive cells and ␤ cells with depleted phogrin, insulin activates IR and downstream signaling pathways following transient inactivation of PTP1B by hydrogen peroxide (thin arrows). Prolonged activation of IR by insulin induces irreversible oxidation of PTP that can lead to chronic activation of IRS2 signal transduction (heavy arrows). After a period of time, internal feedback inhibition signals dampen further activation of IRS2 signaling. In this negative feedback system, Akt and/or TORC1 can phosphorylate IRS2 at serine residues to promote its degradation by the proteasome (dashed arrows). B, in pancreatic ␤ cells, insulin secretion stimulated by high glucose (GSIS) triggers autocrine-and paracrine-mediated IR activation. GSIS promotes translocation of phogrin to the plasma membrane, where it can interact with activated IR. Localization of phogrin at the IR signaling platform contributes to downstream IRS2 signal transduction by preventing PTP1B oxidation. Dephosphorylation of IR by PTP1B occurs to prevent excess signaling and protects the IRS2 protein from degradation.

Knockout mice
Construction of the targeting vector, homologous recombination using embryonic stem cells (RENKA), and generation of chimeric mice were achieved by Transgenic Inc. (Kobe, Japan). The targeting vector contains a sequence that includes exon 4 of the phogrin gene and a neomycin resistance gene driven by the pgk promoter (pgk-neo), which are both flanked by the loxP cassette. Homologous recombination replaces the gene with the targeting sequence. Mutant lines were maintained by crossing male and female homozygotes. RIP-cre mice (37) were maintained as heterozygotes by backcrossing with C57Bl/6J mice (Japan SLC). Control (Cre ϩ/Ϫ _Phogrin ϩ/ϩ ) and knockout (Cre ϩ/Ϫ _Phogrin flox/flox ) mice for the experiments were obtained by crossing heterozygotes. Male mice with similar body weights and ages were used for all experiments. Genotype analysis was done by PCR. The primers for Cre were 5Ј-ACC-TGATGGACATGTTCAGGGATCG-3Ј and 5Ј-TCCGGTTA-TTCAACTTGCACCATGC-3Ј, and the primers for phogrin and floxed allele were 5Ј-CCAGCGTGTGATTGCCCAGGAG-3Ј, 5Ј-ATTCGCAGCGCATCGCCTTCTATCGCCTTC-3Ј, and 5Ј-CCTGCTCATTGCAATGTTGTTAAC-3Ј. Mice had free access to water and standard laboratory chow (CE-2; CLEA Japan) in an air-conditioned room with a 12-h light/dark cycle. All animal experiments were approved by the Gunma University Animal Care and Experimentation Committee (permission no. 16-017) and were conducted according to its guidelines.

Immunohistochemistry
Immunohistochemical analyses were performed as described previously (45). Paraffin-embedded pancreas sections were labeled with anti-phogrin antibody (1:1,500 dilution) and detected using an avidin-biotin-peroxidase technique (Vector Laboratories) with hematoxylin counterstaining. Image acquisition was completed using a microscope (Olympus BX-50) equipped with a SenSys TM charge-coupled device camera (Photometrics).

Thymidine incorporation and apoptosis analysis
Mouse pancreatic islets were isolated and cultured as described previously (24). After culturing for 48 h, [ 3 H]thymidine was added at a final concentration of 1 mCi/ml to pools of 50 islets for an additional 24 h. The DNA was precipitated with ice-cold 10% trichloroacetic acid and solubilized in 0.3 N NaOH. Aliquots were counted in scintillation fluid, and thymidine incorporation into islet cell DNA was determined. Isolated islets were dispersed into cells with 0.05% trypsin, 0.02% EDTA and cultured on 8-well Lab-Tek chamber slides (Thermo Scientific Nunc) precoated with poly-L-lysine (Sigma). Apoptosis analysis of primary ␤ cells was performed as described previously (47). Briefly, islet cells were fixed with 4% paraformaldehyde and then incubated with anti-insulin followed by Rhodamine Red TM -X-conjugated, anti-guinea pig IgG secondary antibody (Jackson ImmunoResearch) to detect ␤ cells. Condensation or fragmentation of nuclei (apoptotic characteristics) was assessed by 4Ј,6-diamidino-2-phenylindole staining. Insulin-positive islet ␤ cells having normal or apoptotic nuclei were counted, and the cell death rate was calculated. At least 10 different fields from each mouse sample were randomly selected to count at least 100 ␤ cells for calculation of apoptosis rates.

Phogrin connects insulin secretion and signaling
for oxidized PTP; 1:4,000 for IRS2; 1:2,000 for Rab27; 1:1,000 for GFP; 1:800 for PTP1B; 1:5,000 for VAMP2). Band density was measured by densitometry, quantified using gel-plotting macros of the NIH Image version 1.62 program, and normalized to an indicated sample in the same membrane.

Semiquantitative RT-PCR
RT-PCR analysis was performed as described previously (24). Amplified signals stained with ethidium bromide were quantified with an ATTO Cool Saver system (ATTO, Tokyo, Japan).

In vitro assay with GST proteins
An in vitro binding assay (29) and dephosphorylation assay (49) were combined. PTP1B and TCPTP cDNAs were subcloned into the pGEX6P-1 vector. Bacterially expressed GSTfused proteins were then affinity-purified with glutathione-Sepharose beads and eluted with reduced glutathione or incubated with PreScission protease (GE Healthcare). Purified proteins were dialyzed with 10 mM Tris buffer. COS7 cells expressing IR-EGFP were treated with 100 nM insulin for 10 min and then extracted with lysis buffer A. IR-EGFP was immunoprecipitated with agarose-conjugated anti-GFP (RQ2, MBL Co.) and washed with PTP buffer (20 mM Tris, pH 6.8, 150 mM NaCl, 2 mM EDTA, 25 mg/ml BSA, and 1 mM dithiothreitol) containing 0.05% Nonidet P-40. IR-EGFP immobilized on agarose beads were incubated at 25°C with 2 pmol of each GST protein and 1 pmol of recombinant PTP1B in 0.2 ml of PTP buffer for 20 min. The beads were washed three times, and the bound proteins were analyzed by immunoblotting. Each purified GST protein (4 pmol) was preincubated with or without recombinant PTP1B for 10 min. PTP activity was then measured with pNPP as a substrate in a buffer containing 20 mM MES, pH 6.0, 2 mM EDTA, and 10 mM pNPP. The reaction was terminated with NaOH, and absorbance was measured at 410 nm.

Fluorescence microscopy
MIN6 and hepatocytes were incubated in the dark at 37°C for 5 min with 5 M CM-H 2 DCFDA (Invitrogen) on poly-L-lysine (Sigma)-coated coverslips. The cells were mounted on a glass slide with Hanks' buffer. DCF fluorescence was observed with an epifluorescence microscope (BX-50) as described previously (50). To avoid photooxidation of the indicator dye, the fluorescence images were collected by a single rapid scan with identical parameters (e.g. contrast and brightness) for all samples.

Statistical analysis
Results are given as the mean Ϯ S.E., except when otherwise indicated. Differences between groups were analyzed using Student's t test or the Mann-Whitney U test. p values Ͻ 0.05 were considered statistically significant.