Insulin Rescues Retinal Neurons from Apoptosis by a Phosphatidylinositol 3-Kinase/Akt-mediated Mechanism That Reduces the Activation of Caspase-3*

The ability of insulin to protect neurons from apoptosis was examined in differentiated R28 cells, a neural cell line derived from the neonatal rat retina. Apoptosis was induced by serum deprivation, and the number of pyknotic cells was counted. p53 and Akt were examined by immunoblotting after serum deprivation and insulin treatment, and caspase-3 activation was examined by immunocytochemistry. Serum deprivation for 24 h caused (cid:1) 20% of R28 cells to undergo apoptosis, detected by both pyknosis and activation of caspase-3. 10 n M in- sulin maximally reduced the amount of apoptosis with a similar potency as 1.3 n M (10 ng/ml) insulin-like growth factor 1, which acted as a positive control. Insulin induced serine phosphorylation of Akt, through the phosphatidylinositol (PI) 3-kinase pathway. Inhibition of PI 3-kinase with wortmannin or LY294002 blocked the ability of insulin to rescue the cells from apoptosis. SN50, a peptide inhibitor of NF- (cid:1) B nuclear translocation, blocked the rescue effect of insulin, but neither insulin

the growth factors that promote survival and growth of primary retinal ganglion cells in culture.
The insulin receptor has multiple downstream targets that are conserved in many cell types. It is generally thought that its antiapoptotic mechanism signals primarily through PI 3kinase to activate the serine-threonine kinase, Akt (5). Akt inhibits many proapoptotic targets by phosphorylating caspase-9, glycogen synthase kinase, BCL-2-associated death promoter, and members of the forkhead family (6 -12). Insulin also inhibits the release of cytochrome c from mitochondria and IB, which regulates NF-B translocation (13)(14)(15)(16), similar to the IGF-1 1 receptor-stimulated pathway (17,18).
NF-B is a transcription factor that is translocated from the cytoplasm to the nucleus and has been implicated in neuronal survival (19). NF-B may also be involved in insulin signaling to prevent apoptosis. Insulin activates NF-B in Chinese hamster ovary cells transfected with the insulin receptor, and the anti-apoptotic signal of insulin is blocked by overexpression of a dominant negative IB-␣, which is the inhibitory peptide for NF-B (20). It is possible that the anti-apoptotic effect mediated by NF-B is due to increased transcription of manganesesuperoxide dismutase (21). Overexpression of the NF-B c-rel subunit in neurons blocked apoptosis as potently as IGF-1, whereas overexpression of a dominant-negative IB-␣ enhanced apoptosis, suggesting that the function of NF-B in neurons is primarily neuroprotective (22).
The tumor suppressor gene p53 is also strongly implicated in apoptosis of neurons (23,24). Although p53 is known to induce growth arrest, it is likely that it induces apoptosis by an independent mechanism (25). Overexpression of p53 in postmitotic neurons leads to apoptosis (26) and involves activation of BAX and caspase-3 (27). p53 also up-regulates glyceraldehyde-3phosphate dehydrogenase in cerebellar granule neurons during apoptosis (28). However, p53 may not be involved in apoptosis induced by factors that do not lead directly to DNA damage, such as low potassium (29,30). The involvement of p53 in apoptosis of retinal neurons is unclear, but it is elevated after retinal ischemia (31).
Here, we tested the hypothesis that physiological concentrations of insulin may act as a survival factor in R28 cells, which are a subclone of E1A-NR.3 transfected rat retina cells (32,33). R28 cells were originally isolated from neonatal (P12) rat retina as a mixed cell type and transfected with an E1A construct (34). They have been described as having both neuronal and glial characteristics, expressing the neuron-specific antigen Thy 1.1 and some glial cell markers (34,35). We first show that a combination of laminin substrate and cAMP induces a neuronal phenotype in these cells (36). The ability of insulin to rescue R28 cells from apoptosis induced by serum deprivation was investigated. The data demonstrate that the PI 3-kinase/ Akt pathway mediates the survival effect of insulin and leads to inhibition of caspase-3.

EXPERIMENTAL PROCEDURES
Materials and Reagents-Insulin was from bovine pancreas extract (Sigma, St. Louis, MO) 28 USP units/mg; recombinant human IGF-1 (R3 mutant, resistant to IGF-binding proteins) was from Upstate Biotechnology (Lake Placid, NY); cell culture medium and bovine newborn serum were from Life Technologies, Inc. (Rockville, MD). All dry chemicals were from Fisher and Sigma. Wortmannin, LY294002, PD98059, and SN-50, an NF-B peptide translocation sequence inhibitor, were from Biomol. Laminin and RNase were from Sigma. The IGF-1 receptor inhibitor JB3 was a generous gift from Dr. Lois E. H. Smith, Boston Children's Hospital, Harvard University.
Cell Culture and Protein Extraction-R28 Cells were seeded at 4 ϫ 10 5 cells/cm 2 on 60-mm dishes coated with laminin and 2 ϫ 10 5 cells/ cm 2 on glass coverslips for immunocytochemistry and apoptosis assays. They were fed with Dulbecco's modified Eagle's medium supplemented with 10% newborn bovine serum and 250 M pCPT-cyclic AMP, a cell-permeable cAMP analogue (Sigma).
At the start of each experiment, all cells were rinsed in PBS and then treated with fresh medium with or without serum, insulin, recombinant mutant IGF-1, or IGF-1 receptor inhibitor, JB3 (37). The half-life of insulin in this medium was 49.7 h, measured by insulin enzyme-linked immunosorbent assay (Alpco, Windham, NH) in samples of culture medium taken from cells treated with 10 nM insulin for 48 h. The concentration of insulin in the undiluted serum was estimated as 0.2 ng/ml by enzyme-linked immunosorbent assay.
Quantification of Apoptosis-Cells were grown to 90% confluence before 24-h experimental treatment. The coverslips were rinsed in PBS and fixed in 1% paraformaldehyde before staining with 1 g/ml propidium iodide and 0.5 mg/ml DNase-free RNase A (Sigma). Cells were viewed using the ϫ 40 objective of an Olympus BH-2 fluorescence microscope mounted with a Sony 3CCD video camera attached to an IBM personal computer running image analysis software (Optimus, Media Cybernetics, Silver Spring, MD). Five visual fields were randomly sampled from each coverslip and all the cells in each field were counted as either pyknotic, mitotic, or viable. The number of pyknotic cells was summated in the five sampled regions and expressed as percentage of apoptosis per coverslip ϭ (total number of pyknotic cells) (total number of pyknotic cells ϩ total number of viable cells) ϫ 100. A similar approach was used to estimate the percentage of mitotic cells.
Immunocytochemistry-R28 cells grown on laminin-coated coverslips were deprived of serum and treated for 24 h with or without 10 nM insulin or 1.3 nM IGF-1-3R, which has decreased affinity for IGFbinding proteins relative to native IGF-1 (39). The cells were fixed in 1% paraformaldehyde and then incubated with a rabbit polyclonal antibody against activated caspase-3 (CM-1, Idun Pharmaceuticals, La Jolla, CA) or a rabbit anti-rat antibody against p65 of NF-B (a generous gift of Dr. Shao-Cong Sun, Pennsylvania State University) (both antibodies at 1:1000). The secondary antibody was rhodamine red X-conjugated donkey anti-rabbit (1:2000, Jackson Immunologicals, West Grove, PA). Cells were simultaneously stained with the nuclear dye bisbenzimide (Hoechst dye 33258, 0.5 g/ml, Sigma).
Transfections-R28 cells were transfected using LipofectAMINE PLUS TM reagent (Life Technologies, Inc.) with 2.5 g of either the parental pCMV4 or the pCMV4-HA-IB␣ S36A expression vector and seeded on coverslips. The parental vector pCMV4 and the cDNA expression vector encoding the hemagglutinin (HA)-tagged IB␣ with a serine to alanine substitution at residue 36 (S36A) (pCMV4-HA-IB␣ S36A) were generous gifts from Dr. Shao-Cang Sun from Department of Microbiology and Immunology, Pennsylvania State University College of Medicine (Hershey, PA) (40).
24 h after transfection, the cells were either left in serum or deprived of serum with or without 10 nM insulin and incubated for another 24 h. The cells were fixed with 1% paraformaldehyde and blocked at room temperature for 1 h in PBS containing 0.1% Triton and 10% donkey serum (PBST). They were then incubated at room temperature for 1 h in PBST with rabbit anti-HA polyclonal antibody (1:200, CLONTECH, Palo Alto, CA). The cells were washed and incubated at room temperature for 1 h with rhodamine red X-conjugated anti-rabbit IgG (1:2000) and 0.5 g/ml Hoechst 33258.
Statistical Analysis-Statistical comparisons were made by one-way analysis of variance with post hoc Student-Newman-Keuls multiple comparisons test (Instat 2.0, GraphPad Software). Statistical significance was accepted if p Ͻ 0.05.

Induction of a Neuronal Phenotype in R28
Cells-R28 cells have been described as expressing both neuronal and glial cell markers (34). In order to induce an entirely neuronal phenotype, the cells were seeded on laminin and fed with medium containing cell-permeable cAMP. Under these conditions, the R28 cells adopt a neuronal morphology, with neurite-like projections ( Fig. 1, A and B), and their rate of division is reduced. Immunoblotting showed that the differentiated R28 cells contain the neuronal markers, neuron-specific enolase (Fig. 1C), and 200-kDa neurofilament (both phosphorylated and nonphosphorylated; Fig. 1D) but do not have detectable levels of glial fibrillary acidic protein ( Fig. 1E; no band could be detected when the brightness and contrast settings on the blot image were increased to the upper limits of sensitivity). Immunoreactivity for neuron-specific enolase was also abundant, measured by immunocytochemistry, and distributed homogeneously throughout the entire cell population (data not shown). Immunocytochemistry also revealed immunoreactivity for phosphorylated neurofilament but not glial fibrillary acidic protein or the endothelial cell antigen von Willebrand factor (data not shown). Thus, R28 cells can be differentiated in this way to provide a useful in vitro model of retinal neurons.
Insulin Is a Survival Factor for Retinal Neurons-Insulin mediates cell survival in primary cell cultures of cerebellar neurons and retinal ganglion cells (1,4). To determine whether insulin can also act as a survival factor for the R28 model of retinal neurons, the effect of insulin on apoptosis in serumdeprived cells was quantified. A time course study showed that the rate of apoptosis peaked at 16 -18 h after serum withdrawal (data not shown). Cells were deprived of serum for 24 h and treated with either 10 nM insulin or 1.3 nM recombinant mutant IGF-1 (3R), which served as a positive control. The number of apoptotic cells was counted after propidium iodide staining. Between 10 and 20% of cells were pyknotic after 24 h of serum deprivation, whereas less than 1% were pyknotic in control conditions ( Fig. 2A). Treatment with insulin significantly reduced the frequency of pyknosis ( Fig. 2B; p Ͻ 0.05). Higher concentrations of insulin had no greater effect on apoptosis, indicating that the maximal effect was at 10 nM ( Fig.  2C). Insulin treatment had no effect on the frequency of mitosis (Fig. 2C). The IGF-1 inhibitor JB3 (37) prevented IGF-1 from rescuing cells by 50% but did not block the survival effect of insulin (data not shown). There was no greater effect of adding insulin and IGF-1 together (data not shown). These results suggest that insulin blocks apoptosis in retinal neurons by increasing cell survival rather than proliferation. Differentiated R28 Cells Express Functional Insulin and IGF-1 Receptors-The insulin receptor signaling system in R28 cells was explored by immunoblotting for the ␣-subunit of the insulin receptor. R28 cells express both the 125-kDa isoform of the ␣-chain, similar to the most abundant form in liver (41), as well as the 115-kDa isoform that is typical of the adult retina (Fig. 3A). By immunoprecipitation of the ␤-subunits of the insulin or IGF-1 receptors, followed by phosphotyrosine immu- B, the number of pyknotic cells was measured as the percentage of pyknotic nuclei out of the total number of cells counted (average from five randomly sampled visual fields in three coverslips). There were almost no pyknotic cells in the control cultures. Addition of 10 nM insulin to control cells had no effect on pyknosis. Serum withdrawal significantly increased the number of pyknotic cells (*, p Ͻ 0.001 compared with all other groups). Insulin significantly reduced the number of pyknotic cells compared with the untreated serum deprived cultures (*, p Ͻ 0.001; n ϭ 3 coverslips). 1.3 nM IGF-1reduced apoptosis to a similar degree. C, rescue by insulin was dose-dependent. 1 nM insulin did not significantly reduce the number of pyknotic cells compared with the untreated serum deprived cultures, whereas 10 nM insulin significantly reduced the number of pyknotic cells (**, p Ͻ 0.001; n ϭ 3 coverslips). No further reduction in the number of pyknotic cells was observed with 50 or 100 nM insulin. The number of mitotic bodies was also counted and expressed as a percentage of the total number of cells. Insulin treatment did not alter the number of mitotic cells in any of the culture conditions. n.s., no significant difference. All comparisons were by analysis of variance and the Student-Newman-Keuls test.
noblotting, it was determined that the insulin receptors were tyrosine-phosphorylated by 10 nM insulin, whereas the IGF-1 receptors were not (Fig. 3B). By two-point binding assay with 0.5 nmol of 125 I-bovine insulin, the concentration of insulin receptors in the R28 cells was 110 pmol/mg total protein, which is comparable to that measured in liver plasma membranes and hepatocytes (42). The content of IGF-1 receptors in R28 cells was also similar to that of whole retina, measured by immunoblot. The content of insulin receptors in R28 cells was also comparable to that of equivalent protein samples of whole retina and liver lysates (data not shown). These data show that the survival effect of 10 nM insulin is mediated through the insulin receptor and not IGF-1.
Retinal Cell Survival Is Mediated by PI 3-kinase and Akt-Many growth factors have antiapoptotic effects via activation of PI 3-kinase and Akt/protein kinase B in various cell types (6), so the involvement of these pathways was investigated in differentiated R28 cells. Cells were treated with the PI 3-kinase inhibitors wortmannin (100 nM) or LY294002 (10 M) or with the MEK kinase inhibitor PD98059 (25 M), dissolved in dilute Me 2 SO. Inhibition of PI 3-kinase activity with wortmannin (100 nM) or the more specific PI 3-kinase inhibitor LY294002 (10 M) blocked the ability of both insulin and IGF-1 to rescue R28 cells from apoptosis induced by serum deprivation (Fig. 4). The MEK1 inhibitor PD98059 also blocked insulin-mediated survival, but the magnitude of inhibition was not as great as that induced by the PI 3-kinase inhibitors. Therefore, the PI 3-kinase pathway has a primary role in survival mediated by insulin.
Activation of the downstream target of the PI 3-kinase pathway, Akt, was investigated by immunoblotting with antibodies to total Akt and to Ser 473-phosphorylated Akt. Total Akt content was unchanged after serum deprivation, and there was a low basal level of Akt phosphorylation. The amount of Ser 473-phosphorylated Akt increased significantly within 10 min of insulin stimulation in the absence of serum (Fig. 5A). Akt phosphorylation was also induced by insulin in cells grown with serum (data not shown). Wortmannin and LY294002 both blocked the Akt serine phosphorylation induced by insulin when serum was removed (Fig. 5B). The reduction in Akt phosphorylation was statistically significant (Fig. 5C; p Ͻ 0.05). Similar results were obtained after stimulation with IGF-1. These data suggest that insulin induces Akt phosphorylation through PI 3-kinase in retinal neurons. with insulin or IGF-1, and the ␤-subunits of the insulin and IGF-1 receptor were immunoprecipitated. Samples were resolved in parallel by SDS-polyacrylamide gel electrophoresis followed by phosphotyrosine immunoblotting to determine autophosphorylation of the respective receptors. Insulin added for 2 min induced tyrosine phosphorylation of the insulin receptor but not the IGF-1 receptor. IGF-1 treatment for 2 min also induced some weak phosphorylation of the insulin receptor and strongly phosphorylated the IGF-1 receptor. There was no detectable basal phosphorylation of either receptor. Whereas phosphorylation of the insulin receptor was reduced after 24 h of insulin treatment, phosphorylation of the IGF-1 receptor was more attenuated. Immunoblots are typical of three replications.

FIG. 4. PI 3-kinase pathway plays a dominant role in neuroprotection mediated by insulin and IGF-1. Differentiated R28 cells
were serum-deprived and treated with the PI 3-kinase inhibitors wortmannin or LY294002, with the MEK1 inhibitor PD098059, or with 0.53% Me 2 SO for 30 min, followed by addition of 10 nM insulin or 1.3 nM IGF-1. The cells were stained with propidium iodide after 24 h of treatment, and the pyknotic nuclei were counted. A, serum deprivation significantly elevated the number of pyknotic cells, as before. Addition of the PI 3-kinase inhibitors, the MEK inhibitor, or Me 2 SO vehicle in the absence of serum did not alter the number of pyknotic cells. B, insulin significantly reduced the number of pyknotic cells compared with untreated serum-deprived cells. The effect of insulin was reversed by wortmannin and LY294002, but PD098059 did not block the significant reduction in pyknotic cells caused by insulin treatment. C, IGF-1 also reduced the number of pyknotic cells compared with untreated serum-deprived cells; this effect was reversed by wortmannin and LY294002 but not PD098059. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; n.s., not significant. n ϭ 3 coverslips per experiment).
Insulin Inhibits Caspase-3 Activation-Caspase-3 is a cysteine protease that becomes active during the later stages of apoptosis. Previous studies have shown that activated caspase-3 can be detected by immunocytochemistry using the antibody CM-1 (43). Immunocytochemistry for activated caspase-3 was performed on serum-deprived R28 cells using the CM-1 antibody. The cells were counterstained with Hoechst 33258 dye to reveal nuclear morphology. In cells grown with serum, there was little CM-1 immunoreactivity, and almost all cells had normal nuclear morphology (Fig. 6). In contrast, many of the serum-deprived cells were immunoreactive for caspase-3. Those cells with caspase-3 immunoreactivity also tended to have pyknotic nuclei. Both insulin and IGF-1 reduced CM-1 immunoreactivity and the number of pyknotic cells, whereas LY294002 blocked these effects. Thus, insulin inhibits the activation of caspase-3 by a PI 3-kinase-dependent mechanism.
Insulin Signaling and NF-B-Insulin induces activation of the nuclear transcription factor NF-B in sympathetic and cerebellar neurons (22,44,45). In order to test whether translocation of p65 NF-B is also required for rescue by insulin in R28 cells, four approaches were used. First, the peptide inhibitor of the NF-B nuclear localization sequence, SN50, was tested. SN50 is a cell-permeable antisense peptide that binds to the nuclear translocation region of NF-B and prevents it from being transported to the nucleus (46). SN50 alone increased the number of pyknotic cells in serum-deprived cultures (Fig. 7A), whereas the control peptide, SN50M, which is mutated to prevent its peptide binding ability, had no effect on apoptosis. SN50 blocked the ability of insulin to rescue R28 cells from apoptosis and also increased the number of pyknotic cells sig-nificantly more than serum deprivation (Fig. 7B). In contrast, SN50 blocked the ability of IGF-1 to prevent apoptosis but did not increase the amount of apoptosis any more than serum deprivation (Fig. 7C). The mutated peptide, SN50M, did not block the effects of insulin or IGF-1. These data suggest that NF-B translocation may be involved in cell survival induced by insulin.
Second, to examine nuclear translocation of the p65 subunit of NF-B directly, we used immunocytochemistry in R28 cells treated with or without insulin. R28 cells were grown in the presence of insulin for 6 or 24 h, fixed, and labeled with an antibody to the p65 subunit of NF-B. No nuclear localization of p65 was observed after any of the treatment conditions (Fig.  8), suggesting that insulin does not induce nuclear translocation of NF-␤ in R28 cells.
Third, R28 cells were transfected with a dominant negative construct of IB, which cannot be activated by phosphorylation. About 20% of the cells were transfected successfully, indicated by positive HA immunoreactivity. However, none of the transfected cells contained pyknotic nuclei or caspase-3 immunoreactivity after serum deprivation (data not shown). These data suggest that transfection of a dominant negative IB caused the cells to be resistant to apoptosis induced by serum deprivation.
In the fourth and final approach, we examined the phosphorylation of IB. In order for NF-B to be translocated to the nucleus, IB must be phosphorylated. To test whether this occurs during apoptosis in R28 cells, immunoblotting with a phospho-specific antibody to IB was carried out on lysates of cells that had been serum-deprived and treated with insulin, IGF-1, or LY294002. The cell lysates were also immunoblotted for Ser 473-phosphorylated Akt and compared with lysates of HeLa cells as positive controls. Compared with HeLa cells, there was little IB phosphorylation in R28 cells deprived of serum for 24 h. Treatment with insulin, IGF-1, or LY294002 also had little effect on IB phosphorylation (Fig. 9). These data suggest that IB phosphorylation does not occur in R28 cells under these conditions. In total, the data from these four experiments suggest that NF-B does not mediate the rescue effect of insulin and that SN50 may block insulin signaling by an unknown mechanism in R28 cells.
Insulin Does Not Induce Survival by Altering p53 Expression-The tumor suppressor protein p53 has been implicated as a mediator of apoptosis in some cell types, but its involvement is determined by the type of apoptotic stimulus (30,47). The PI 3-kinase pathway can block p53-mediated apoptosis (48). To examine its involvement in apoptosis of retinal neurons, p53 was quantified by immunoblotting R28 cells after serum deprivation with or without insulin (Fig. 10A). There was a transient increase in p53 content that reached a peak after 2-4 h of serum deprivation, which fell to almost undetectable levels after 24 h (Fig. 10B). Treatment with 10 nM insulin did not alter the transient elevation of p53 in serumdeprived cells (Fig. 10C). Therefore, insulin may induce survival of R28 cells by affecting a target either downstream or independent of p53.

DISCUSSION
The data presented here demonstrate that physiological concentrations of insulin can protect retinal neurons from apoptosis induced by serum deprivation in culture. This concept has been demonstrated in in vitro models of neurons from other parts of the nervous system, including sensory nerves (49), cortex (50), cerebellar granule cells (51), spinal motor neurons (52), and PC12 cells (53). In this study, we used differentiated R28 cells to model retinal neurons, and the results further support previous observations that both insulin and IGF-1 are trophic factors for the survival of retinal neurons (4). To the best of our knowledge, this is the first study to examine the FIG. 7. Blocking the NF-B nuclear translocation sequence inhibits the rescue induced by insulin. R28 cells were serumdeprived for 24 h in the presence of SN50 (75 nM), a peptide blocker of the nuclear translocation sequence in NF-B, or SN50M, a mutated peptide incapable of blocking the nuclear translocation sequence. The number of pyknotic cells was counted as before. A, serum deprivation increased the number of pyknotic cells, as before. SN50 significantly increased the number of pyknotic cells beyond serum deprivation alone, whereas SN50M had no effect. B, 10 nM insulin reduced the number of pyknotic cells induced by serum deprivation, as before. SN50 prevented the reduction induced by insulin and significantly elevated the number of pyknotic cells compared with serum deprivation alone, whereas SN50M had no effect on the reduction in pyknotic cells induced by insulin. C, IGF-1 also reduced the number of pyknotic cells induced by serum deprivation. SN50 blocked the effect of IGF-1 but did not elevate the number of pyknotic cells beyond that of serum deprivation alone. SN50M did not change the effect of IGF-1. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; n.s., not significant. All statistical comparisons were relative to the serum-deprived group; n ϭ 3 coverslips per experiment. signaling mechanisms for retinal neurons in response to insulin. The potency of insulin to act as a survival factor in vitro has led to the suggestion that insulin alone is sufficient for neuronal survival, at least at high concentration (1). However, high concentrations of insulin can also activate the IGF-1 receptor. In the present study, lower concentrations of insulin were used. Insulin had a maximal effect in reducing apoptosis at a concentration of 10 nM, which is within the physiological range for plasma insulin and does not stimulate the IGF-1 receptor. Therefore, we conclude that the protective effect of insulin was specific for the insulin receptor.
Although it is not likely that neurons of the retina are directly exposed to plasma when the blood-retinal barrier is intact, the concentration of insulin in the ocular fluids-vitreous and aqueous humors-has been estimated to be in the physiological range (54,55). Insulin may be transported into the retina across the blood-retinal barrier similar to the bloodbrain barrier (56,57) or may arise by de novo synthesis within the retina, as indicated by the presence of preproinsulin mRNA in retina (58). Although the source of insulin is not known, it is clear that both insulin and its receptor are abundant in the retina, especially in the inner plexiform layer, which is mostly formed by neuronal projections (3,59,60). There is strong evidence that insulin has biological activity in the central nervous system. The insulin receptor is expressed in the brain (61). Insulin-like peptides may also be released from cultured neuronal cells, synaptosomes, and astrocytes (62)(63)(64). Therefore, it is reasonable to suggest that insulin signaling in retinal neurons is active in vivo with canonical physiological consequences.
The R28 cells used here as a model of retinal neurons were derived from mixed retinal cells but have been used successfully in other studies of neuronal apoptosis (34,65). Under the culture conditions employed here, these cells had a neuronal phenotype with no evidence to suggest glial or endothelial cell characteristics. When laminin is used as a substrate for other neuronal cell lines, such as PC12 cells, it causes them to develop extensive neurite outgrowth (66). This effect may be modulated by protein kinase C, but a direct regulatory role has not been established. The role of cAMP in further differentiating R28 cells is also important to consider. cAMP elevation promotes survival of spinal motor neurons in vitro (67), possibly by elevating the recruitment of the neurotrophin TrkB receptor to the plasma membrane (68). It is not clear whether cAMP has a similar effect on R28 cells. Although the E1A transfection of R28 cells causes their persistent proliferation, we have shown that the addition of laminin and cAMP provides a more stable model of retinal neurons. The characteristics of the insulin receptor in the retina differ from those seen in peripheral tissues such as liver, muscle, and fat. The predominant insulin receptor in the central nervous system is less glycosylated on both ␣ and ␤ subunits (41). The differentiated R28 cells express both the 125-and 115-kDa species of the ␣-subunit of the insulin receptor. The ␤-subunits of the insulin receptors were tyrosine-phosphorylated by 10 nM insulin, but the IGF-1 receptor was not. Thus, differentiated R28 cells respond to insulin stimulation with receptor tyrosine phosphorylation. This is in agreement with other data showing that exogenous physiological concentrations of insulin also cause receptor tyrosine phosphorylation in intact retinas (69).
Insulin reduced the activation of caspase-3 in R28 retinal neurons, and this effect was blocked by the PI 3-kinase inhibitors. These data imply that insulin regulates the activation of caspase-3 through the PI 3-kinase/Akt pathway. Akt can directly phosphorylate caspase-9, which is at the head of the caspase protease cascade (7). Therefore, caspase-3 activation may be blocked by insulin receptor signaling in retinal neurons through Akt phosphorylation of caspase-9. This implies that reduction of insulin stimulation could lead to apoptosis mediated by caspase-3 activation in retinal neurons.
Previous reports have suggested that the PI 3-kinase pathway activates NF-B in response to insulin, to promote neuronal survival. We used four approaches to test this hypothesis in R28 cells. First, the peptide inhibitor of p65 nuclear translocation, SN50, blocked the protective effect of insulin. These data suggested that translocation of NF-B to the nucleus is required for insulin to rescue the R28 cells. However, additional experiments did not support this initial interpretation. Transfection of a dominant negative IB mutant blocked the apoptosis induced by serum deprivation, rendering this approach inoperable. Therefore, effects of insulin on IB phosphorylation were examined. Serum deprivation or insulin did not induce detectable IB phosphorylation despite a relative abundance of IB protein compared with HeLa cells. Because phosphorylation of IB is generally considered to be required for NF-B translocation, it appears that this may not be an important insulin signaling mechanism in R28 cells. Finally, immunohistochemistry revealed no nuclear translocation of the p65 subunit in response to insulin. Although SN50 augmented apoptosis, it may not be specific for NF-B, because it also blocks other transcription factors with nuclear localization sequences by competing for binding to components of the nuclear transport machinery (46,70). Taken together, these data suggest that another transcription factor requiring nuclear translocation may be involved. Therefore, the rescue effect of insulin in R28 cells deprived of serum may not occur by activation of NF-B.
FIG. 10. p53 is transiently activated by serum deprivation and is not affected by insulin. R28 cells were serum-deprived for between 1 and 24 h and immunoblotted for p53. In a similar experiment, the cells were serum-deprived and treated with 10 nM insulin. A, p53 peptide resolved as a single band in equal protein-loaded samples. B, quantification relative to 15 ng of p53 peptide showed that the total content of p53 was transiently elevated after 3 h of serum deprivation. By 24 h, the total content of p53 was diminished to less than control levels. C, 10 nM insulin did not alter the change in p53 expression in serum-deprived cells.
In conclusion, our data show that insulin can act as a survival factor for R28 cells by stimulating the PI 3-kinase, Akt/ protein kinase B pathway and inhibiting the activation of caspase-3. Although p53 is transiently invoked during this form of cell death, the survival effect of insulin does not alter p53 expression, implying that it modulates downstream effectors of p53. Insulin may regulate multiple systems in retinal tissue, and it is likely that part of its function is to act as a survival signal for retinal neurons. Recent evidence shows that experimental diabetes induces retinal neurodegeneration by increasing apoptosis of inner retinal neurons (71). Apoptosis increases soon after the onset of streptozotocin diabetes and is reversed by exogenous insulin. This cell death gives rise to a chronic neurodegeneration in which ϳ10% of retinal ganglion cells is lost after 7.5 months. Insulin also increases glutamine synthetase activity in the retinas of diabetic rats (72)(73)(74). Acute insulin treatment also reverses the effects of diabetes on the expression of glial fibrillary acidic protein and the tight junction protein, occludin, in diabetic rats (75). Taken together these findings raise the possibility that defective insulin signaling could contribute to retinal neurodegeneration in diabetes.