IGF-I mRNA and Signaling in the Diabetic Retina

Human eyes and specimens. Human eyes were obtained from certified eye banks through the National Disease Research Interchange. Criteria for inclusion in the study were age <70 years, the fewest possible chronic pathologies other than diabetes, absence of retinal or hematological diseases and uremia, and absent administration of chemotherapy or life-support measures. Diabetes duration was <15 years to address mostly background retinopathy. The eyes of the diabetic and age-matched nondiabetic donors were enucleated 3 ± 2 and 3 ± 1 h after death, respectively. Eyes or poles were kept at 4°C by the eye bank, and shipped to the laboratory on ice; the time elapsed from death to processing was 28 ± 8 h for the diabetic specimens and 28 ± 6 h for the control specimens.

The eyes used for isolation of total retinal RNA were obtained from six diabetic donors (age 64 ± 8 years, diabetes duration 7 ± 5 years) and six nondiabetic donors (age 62 ± 7 years). The two retinas from each donor were immediately dissected, separated from the retinal pigmented epithelium, pooled, and homogenized in guanidine isothiocyanate. RNA was isolated by centrifugation on cesium chloride followed by ethanol precipitation.

The eyes used for preparation of retinal protein were obtained from 10 diabetic (age 66 ± 7 years, diabetes duration 8 ± 4 years) and 10 nondiabetic donors (age 66 ± 6 years). The dissected retinas were homogenized in ice-cold lysis buffer containing phosphatase and protease inhibitors (30 mmol/l Tris-HCl, pH 7.4, 10 mmol/l EGTA, 5 mmol/l EDTA, 250 mmol/l sucrose, 1% Triton-X-100, 1 mmol/l NaF, 1 mmol/l Na₃VO₄, 1 mmol/l phenylmethylsulfonyl fluoride, 5 μg/ml leupeptin, 5 μg/ml pepstatin, and 15 μg/ml aprotinin). The homogenate was sonicated three times for 2 s, centrifuged at 16,000g for 15 min at 4°C, and the supernatant was collected and stored in aliquots at -80°C. Protein concentration was determined with the Bradford method using bovine serum albumin (BSA) as standard (protein assay kit; Bio-Rad, Hercules, CA). Protein lysates were also prepared from human retinal microvessels isolated from the neural retina by osmotic lysis (3). Homogenates of human brain, heart, and kidney (Protein Medley) were purchased from Clontech (Palo Alto, CA).

Animals and specimens. Sprague-Dawley male rats (6 weeks old [Taconic Farms, Germantown, NY]) were randomly assigned to one of the following groups: normal, diabetic, or galactosemic. Diabetes was induced by intravenous administration of streptozotocin (55 mg/kg body wt, dissolved in citrate buffer, pH 4.5). Experimental galactosemia was induced by feeding regular rat diet containing 30% D-galactose (Sigma, St. Louis, MO). The animals were treated in conformity with the Association for Research in Vision and Ophthalmology resolution on treatment of animals in research. All rats had free access to food and water. Body weight was recorded twice weekly in the diabetic rats, and 1-2 U insulin (NPH insulin; Novo Nordisk, Princeton, NJ) was given subcutaneously as needed to avoid weight loss without preventing hyperglycemia and glycosuria. The last insulin injection was given 5.6 ± 3.6 days before the rats were killed. Whole blood for measurement of HbA_1c was obtained by intracardiac sampling at the time of killing. After 2 and 5 months of diabetes or galactose feeding, the animals were killed by CO₂ inhalation. The eyes were immediately removed, and the retains were dissected under a microscope and processed for RNA or protein isolation as described above. The brain, heart, and kidneys were also excised and processed for protein isolation. The eyes of other rats were fixed in 10% buffered formalin, and the retinas were digested with trypsin (1) to isolate the vascular network.

Four normal rats were used to assess the postmortem stability of the IGF-I mRNA and IGF-IR. One eye of each rat was enucleated and processed at the time of sacrifice, the other eye was enucleated 3 h after sacrifice and kept at 4°C for an additional 24 h to match the timing of events reported above for the human specimens. The retinas were then isolated and processed for RNA and protein extraction.

Sprague-Dawley hypophysectomized and sham-operated control rats were obtained from Taconic Farms. The rats were killed 2 weeks after surgery, and the retina, brain, and liver were collected and processed for RNA isolation.

Northern blot analysis. Northern blot analysis was used to study the levels of IGF-I mRNA in human retinas, and was performed as described previously (21). Membranes were hybridized to a full-length human IGF-I cDNA probe (gift of Dr. Peter Rotwein) containing a 780-nucleotide fragment that is present in all types of IGF-I transcripts, and, to control for loading, to a 457-bp human β-actin cDNA (21). Probes were labeled using the Multiprime labeling kit (Amersham Pharmacia Biotech, Piscataway, NJ) and [³²P]dCTP. The hybridization signal was quantitated using a PhosphorImaging system (Molecular Dynamics, Sunnyvale, CA). Data are expressed as IGF-I-to-β-actin ratio. After quantitation, blots were exposed to Kodak AR films, and the autoradiographs used for photography.

Competitive reverse transcriptase-polymerase chain reaction. Competitive reverse transcriptase-polymerase chain reaction (C-RT-PCR) was used to study the levels of IGF-I mRNA in individual rat retinas because of the small amount of RNA available. Specific competitors for IGF-I and the housekeeping gene β-actin were synthesized by polymerase chain reaction (PCR) using hybrid reverse primers as described by Celi et al. (22). The 3′ half of the hybrid reverse primer corresponds to the sequence of the reverse primer, whereas the 5′ half corresponds to a sequence of the target transcript n nucleotides upstream from the reverse primer. Thus, PCR amplification with the forward primer for the transcript of interest in combination with the corresponding hybrid reverse primer will generate a competitor template identical to the endogenous product but for a deletion of n nucleotides. The primers and PCR conditions for IGF-I and β-actin are described in Table 1. A specific competitor for the housekeeping gene cyclophilin was obtained by RT-PCR of a cyclophilin-armored RNA competitor template containing a 25-bp internal deletion (RT Check kit; Ambion, Austin, TX). After amplification, the competitors were resolved by agarose gel electrophoresis, purified using the QIAEX II gel extraction system (Qiagen, Chatsworth, CA), diluted to 100 attomol/ml in TE buffer (10 mmol/l Tris-HCl, pH 7.5, 0.1 mmol/l EDTA) containing 10 mg/ml glycogen, and stored in aliquots at -80°C. The specificity of each PCR product was verified by size and restriction digestion.

Reverse transcriptions were carried out with 1 μg total RNA as described previously (21). Samples from an equal number of diabetic, galactosemic, and control retinas were always processed simultaneously. Equal volumes of each cDNA were added to three different known amounts of the specific competitor, and PCR amplification was performed with the cycle conditions described in Table 1. The PCR products were resolved by agarose gel electrophoresis (4 and 3% NuSieve; FMC Bioproducts, Rockland, ME, for IGF-I and β-actin, respectively, and 3% Metaphor; FMC Bioproducts, for cyclophilin), visualized with GelStar (FMC Bioproducts), and quantitated with the Bio-Rad Gel Documentation System equipped with the QuantityOne software. The amount of the endogenous transcript in each retina was determined for each of the three competitor concentrations as follows: moles of endogenous = (intensity of endogenous/intensity of competitor) × moles of competitor. The three values for each sample were averaged and normalized for the amount of input RNA by dividing the average value by the corresponding value of the housekeeping gene determined in the same way. Because the accuracy of competitor concentration may not be absolute, the data are presented as transcriptto-housekeeping ratio.

The linearity of the PCRs and the appropriate concentration of competitors to be used were determined as described by Vafiadis et al. (23).

Immunoblotting and immunoprecipitation. The IGF-IR levels in human and rat retinas were determined by immunoblotting of retinal lysates performed as described previously (24). Proteins (50 μg) were resolved by electrophoresis on 7.5% SDS-PAGE. Each gel always included an equal number of samples from diabetic and nondiabetic human retinas, or diabetic, galactosemic, and control rat retinas. Proteins were transferred onto nitrocellulose membrane (Bio-Rad) by electroblotting using 25 mmol/l Tris, pH 8.3, 192 mmol/l glycine, 20% methanol buffer containing 0.1% SDS. After blocking overnight at 4°C in 10% nonfat milk in Tris-buffered saline containing 0.05% Tween-20 (TBST), the blots were probed with rabbit polyclonal antibodies against the α or β subunits of the IGF-IR (SantaCruz Biotechnology, Santa Cruz, CA), diluted in TBST, washed, and incubated with horseradish peroxidase—conjugated secondary antibodies. Immunoreactive bands were visualized by enhanced chemiluminescence (SuperSignal West Pico Chemiluminescent Substrate System; Pierce, Rockford, IL). Densitometric quantitation of the autoradiographs was performed with the model 330A computing densitometer (Molecular Dynamics). Data are expressed as densitometric units/micrograms of protein.

Tyrosine phosphorylation of the retinal IGF-IR was detected by immunoprecipitation followed by Western blot analysis. Equal amounts (200 μg) of solubilized protein obtained from individual retinas of diabetic, galactosemic, and control rats, respectively, were incubated in lysis buffer (described above for the homogenization of retinal samples) with anti-IGF-IR β-subunit antibody at 4°C overnight. Lysates of human skin fibroblasts treated with IGF-I (Sigma; 100 ng/ml in serum-free medium for 30 min) were used to positively identify the tyrosine-phosphorylated IGF-IR β-subunit. The immunocomplexes were precipitated by addition of protein-A sepharose (Sigma), and the resulting immunoprecipitates were washed four times with lysis buffer and resuspended by boiling in Laemmli sample buffer supplemented with β-mercaptoethanol. Proteins were resolved by SDS-PAGE and electroblotted onto nitrocellulose membranes. Blots were blocked in 4% BSA-TBST, and tyrosine-phosphorylated proteins were detected by immunoblotting with a mixture of antiphosphotyrosine monoclonal antibodies (clone PY20; Signal Transduction Laboratories, Lexington, KY, and clone 4G10; Upstate Biotechnology, Lake Placid, NY) both diluted 1:5,000. To control for the amount of immunoprecipitated IGF-IR, companion blots, blocked in 10% nonfat milk, were probed with the IGF-IR β-subunit antibodies. IGF-IR phosphorylation was computed as percentage of control, after normalization for the amount of immunoprecipitated IGF-IR.

The levels of Akt phosphorylation were studied by immunoblot using antibodies that detect Akt only when phosphorylated at Ser 473 (Phospho-Akt [Ser 473] Antibody; New England Biolabs, Beverly, MA). To control for the amount of Akt in each retina sample, companion blots were probed with Akt antibodies (New England Biolabs). Immunoreactive proteins were detected and quantitated as above, and Akt phosphorylation was computed as percentage of control, after normalization for the amount of total Akt.

Analysis of DNA fragmentation in retinal vessels. Microvascular cell apoptosis in rat retinal trypsin digests was studied with the terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) assay using the In Situ Cell Death Detection Kit by Boehringer Mannheim Biochemicals (Indianapolis, IN) (1). TUNEL-positive nuclei were examined for chromatin fragmentation and/or presence of apoptotic bodies, and attributed to specific microvascular cells as described previously (1).

Statistical analysis. The data are summarized with the mean ± SD. Statistical analysis was performed with the unpaired t test in the human study and the hypophysectomized rat study; analysis of variance, followed by the Fisher's multiple comparison test, was used for all other rat studies. Linear regression was used to test correlations.

IGF-I expression is decreased in both human and rat diabetic retinas. Northern analysis detected in the human retina a major IGF-I mRNA transcript of the expected molecular mass of 7.5 kb (Fig. 1). As noted in previous studies (21), the intensity of the β-actin signal was similar in the retinas of diabetic and nondiabetic donors (3,174 ± 794 vs. 3,387 ± 1,032 densitometric units, respectively, P = 0.7), and it was thus used as internal standard. In the diabetic retinas, the levels of IGF-I mRNA were significantly decreased compared with the nondiabetic controls (IGF-I—to—β-actin ratio: 0.017 ± 0.010 vs. 0.050 ± 0.021, respectively; P = 0.005). No correlations were found in diabetic or nondiabetic donors between IGF-I mRNA levels and time to the eyes' enucleation or processing, nor with duration of diabetes in diabetic donors. The stability of the IGF-I transcript in the postmortem period was verified in rat retinas. The IGF-I—to—β-actin ratio—measured by C-RT-PCR—in rat retinas processed 27 h after the time the rats were killed was 98 ± 14% of the ratio recorded in retinas processed immediately.

To confirm the observations made in the human retinas and obtain time-course data, we studied the expression of IGF-I in the retina of diabetic and galactose-fed rats after 2 and 5 months of diabetes or galactose feeding. Because the yield of RNA from individual rat retinas was not sufficient to perform Northern blotting, the levels of IGF-I mRNA were measured by C-RT-PCR. The IGF-I primers were designed to amplify both the IGF-Ia and IGF-Ib alternatively spliced forms of the IGF-I transcript; however, only IGF-Ia was detected in rat retina (and brain) using our conditions. The characteristics of the rats are reported in Table 2. After 2 months of diabetes, body weight was significantly lower and HbA_1c levels significantly higher in diabetic than in control rats, but retinal expression of IGF-I was practically identical in the two groups (IGF-I—to—β-actin ratios 0.011 ± 0.011 and 0.011 ± 0.007, respectively) (Fig. 2). In the diabetic as well as the control rats killed at 5 months, body weight was 27% higher than the value recorded in the respective group at 2 months; however, IGF-I mRNA levels were now different. The IGF-I—to—β-actin ratio had increased in the control rats to 0.022 ± 0.007, but remained 0.013 ± 0.006 in the diabetic rats (P < 0.02 vs. agematched controls) (Fig. 2). At both 2 and 5 months, the body weight and HbA_1c levels of galactose-fed rats were intermediate between those of diabetic and control rats (Table 2), and the IGF-I mRNA levels were not significantly different from those of controls (Fig. 2). Because the β-actin mRNA levels showed large variations among animals, the experiments were repeated using cyclophilin as the internal standard to control for input RNA. In the 19 samples tested at the 5-month time point, the cyclophilin mRNA levels correlated with those of β-actin (r = 0.6, P = 0.006) and the IGF-I—to—cyclophilin ratios yielded the same pattern of results as the IGF-I—to—β-actin ratios. Cyclophilin was used in all successive C-RT-PCR assays.

Retinal IGF-I expression is growth-hormone independent. In streptozotocin-induced diabetic rats, growth-hormone levels as well as growth-hormone receptors have been reported to be low (14). To evaluate the relevance of these abnormalities to the decreased retinal IGF-I levels, we tested whether IGF-I synthesis in the retina is regulated by growth hormone. Hypophysectomized and sham-operated control rats were compared vis-à-vis the levels of IGF-I mRNA in the liver, brain, and retina. In the liver, in which IGF-I synthesis is known to be growth-hormone dependent, IGF-I mRNA levels were significantly decreased in the hypophysectomized rats compared with the controls (IGF-I—to—cyclophilin ratio 5.2 ± 1.7 vs. 14.2 ± 4.2, P = 0.02) (Fig. 3). In contrast, in the brain, in which IGF-I levels have been reported to be growth-hormone independent (25), the IGF-I—to—cyclophilin ratio was similar in hypophysectomized (0.2 ± 0.02) and control (0.2 ± 0.06) rats. The retina behaved like the brain, showing IGF-I mRNA levels clearly unaffected by hypophysectomy (IGF-I—to—cyclophilin ratio 1.0 ± 0.3 vs. 0.6 ± 0.2 in control rats) and indicating that retinal IGF-I levels are growth-hormone independent.

The IGF-IR is abundantly expressed in the retina and modestly increased in human diabetes. Immunoblot analysis of human and rat tissues showed that the two IGF-IR subunits are more abundant in the retina than in the brain, heart, or kidney (Fig. 4A). A larger species of ∼ 180 kDa, most likely representing the IGF-IR precursor, was visible in both the retina and brain, and especially prominent in the rat. The human retinal IGF-IR α-subunit (Fig. 4B) was noted to migrate as a doublet, with the smaller-size species migrating in parallel with the brain α-subunit and the larger-size species with the α-subunit of the isolated retinal microvasculature. Because the larger-size band reflects mostly the contribution of IGF-IR in retinal vessels and glia, it indicates an especially prominent representation of IGF-IR on nonneuronal structures in the retina.

In the retinas of 10 diabetic human donors, the levels of the IGF-IR α- and β-subunits (137 ± 22 and 120 ± 23 densitometric units/μg protein, respectively) were greater than in nondiabetic donors (120 ± 24 and 98 ± 19 densitometric units/μg protein, respectively; P = 0.02). The measurements of the α- and β-subunits showed excellent correlation in individual samples (r = 0.8, P < 0.0001) and reproducibility in different assays (coefficient of variation 14 ± 4% for the seven retinal lysates tested in three or more assays). Rat experiments showed that the levels of the retinal IGF-IR are not affected by the postmortem period (the IGF-IR β-subunit levels in retinas processed 27 h after the time the animals were killed were 97% of levels in retinas processed immediately).

In the retinas of the same diabetic and galactosemic rats tested for IGF-I expression, the levels of the IGF-IR α-subunit were not different from those recorded in control rats (77 ± 22 densitometric units/μg protein in diabetic rats, 84 ± 22 in galactosemic rats, and 72 ± 13 in controls at the 2-month time point; and 93 ± 12, 88 ± 17, and 88 ± 8 in the corresponding groups studied at the 5-month time point).

Phosphorylation of IGF-IR and Akt are not decreased in the retina of diabetic rats. We tested whether the decreased levels of endogenous IGF-I observed in diabetic retinas resulted in compromised activation of the IGF-IR. Tyrosine-phosphorylated IGF-IR β-subunit was positively identified in cultured fibroblasts treated with IGF-I and was clearly detected in the normal rat retina (Fig. 5A). Detection was dependent on immediate processing of the tissue, as shown by the complete absence of the band reacting with antiphosphotyrosine antibodies in rat retinas stored for 24 h as well as human postmortem retinas (Fig. 5B). Phosphorylation of the IGF-IR β-subunit in the rat retina was not altered by 5 months of diabetes (94 ± 18% of control values) or galactosemia (82 ± 8% of control) (Fig. 6). To confirm this result at a functional level, we evaluated in the same retinas the phosphorylation status of Akt, a downstream target of IGF-IR signaling. The level of retinal Akt phosphorylation was not altered in the diabetic rats (123 ± 21% of control; P = 0.07) and was slightly increased in the galactosemic rats (132 ± 14% of control; P = 0.02) (Fig. 7).

Reduced retinal IGF-I expression in diabetic rats is not accompanied by capillary cell apoptosis. The studies of IGF-I signaling performed in total retinal tissue might not have informed adequately about IGF-I signaling to discrete cell populations. Retinal capillary cell apoptosis was thus examined, to determine possible consequences of reduced IGF-I synthesis at the level of the microvasculature. The characteristics of the rats used in these experiments were identical to those reported in Table 2 for companion rats in the respective groups. Only an occasional TUNEL-positive cell was observed in the capillaries of control rats (Table 3), in accordance with our previous findings in a different cohort (1). Two or five months of diabetes did not increase the frequency of capillary cell apoptosis; 5 months of galactosemia, however, resulted in a significant number of TUNEL-positive microvascular cells (P = 0.01 vs. age-matched control and diabetic rats).

We examined in this study whether diabetes leads to decreased IGF-I mRNA levels in the retina, as it does in many other tissues (13,14,15,16), and whether decreased IGF-I synthesis generated a proapoptotic environment. Although we observed that the retinal synthesis of this powerful survival factor is decreased in both human and experimental diabetes, we did not detect changes in the activation of its receptor and phosphorylation of a downstream antiapoptotic effector, nor a temporal relationship between decreased IGF-I synthesis and occurrence of retinal microvascular cell apoptosis.

Decreased IGF-I synthesis appears to be an early feature of diabetic retinopathy, but not an immediate counterpart of insulin deficiency, stunted growth, or defective growth-hormone signaling. The eye donors had a short duration of type 2 diabetes. Even if duration had been longer than declared because of the extended asymptomatic period in this type of diabetes, our previous studies had shown in donors with similar characteristics: only incipient retinal microangiopathy and no proliferative changes (1,26). The diabetes-induced suppression of IGF-I synthesis has a degree of selectivity, because the human diabetic retinas with almost undetectable IGF-I mRNA had shown in a previous study unaltered levels of vascular endothelial growth factor mRNA (21). Insulin deficiency is not prominent in type 2 diabetes, but was present to a degree in the streptozotocin-induced diabetic rats that required insulin to prevent catabolism. Yet during the first 2 months of diabetes, which was also the time of a profound growth delay, retinal IGF-I mRNA levels remained similar to control values. The regulation of retinal IGF-I changed, however, between the second and fifth month of study, and the diabetic rats failed to increase IGF-I mRNA despite a percent increase in body weight similar to control rats during this period. Low levels of serum growth-hormone characteristic of aging (27) and streptozotocin-induced diabetes (14) cannot account for the reduced IGF-I synthesis in the diabetic retina, because the diabetic and nondiabetic eye donors were perfectly matched for age, and retinal IGF-I synthesis was found to be growth-hormone independent, akin to brain IGF-I (25). The retina and the brain may have in place unique mechanisms for IGF-I regulation to compensate for the limited transfer of circulating growth hormone across the blood-retinal and blood-brain barriers (28). Such uniqueness will make it more interesting to reconstruct how diabetes reduces endogenous IGF-I synthesis in these organs. A direct contribution of hyperglycemia is currently uncertain. The isolated hyperhexosemia achieved in the galactose-fed rats was insufficient to fully mimic the effects of diabetes, but it was also of a much lesser degree than diabetic hyperglycemia, as indicated by the lower HbA_1c levels in galactosemic as compared with diabetic rats.

To ascertain whether lower retinal IGF-I synthesis in diabetes could be expected to result in decreased signaling, we measured levels of receptor and receptor phosphorylation. The human diabetic retinas showed a small increase in IGF-IR levels when compared with nondiabetic specimens, and no change was observed in the diabetic rat retina, reflecting the inconsistent pattern of IGF-IR variations in diabetic tissues showing decreased expression of IGF-I (13,16). IGF-IR phosphorylation could be measured only in the rat retina, and the levels were not affected by diabetes, suggesting that either the degree of reduction in IGF-I synthesis was insufficient to compromise receptor activation, or that changes had occurred in other modulatory influences. One such change could be increased retinal synthesis of IGF-II, which binds the IGF-IR with slightly lower affinity than IGF-I (13). The observation, however, that IGF-II mRNA levels are substantially decreased in the brain of diabetic rats (29) does not favor this possibility. A role could be played by increased availability of IGF-I and/or IGF-II derived from the circulation. Just as circulating IGFs can cross the capillary barrier in selected areas of the brain (30), they may cross the capillary barrier in the retina; and if this barrier is compromised by diabetes, larger amounts of IGFs may reach the retinal tissue. We could not perform measurements to test this possibility, because our rats were not perfused at the time of killing, and therefore retinal IGFs levels would have reflected an unknown contribution from IGFs in trapped blood. On the other hand, recent evidence suggests that endocrine-acting IGF-I is less important than locally produced IGF-I in tissue signaling and action (31). The activity of the IGF-IR can also be modulated by IGF-I—binding proteins, which can either enhance or reduce IGF-I availability to its receptor (13), and the levels of several of these proteins change in diabetes (13). The difficulty in testing the role of any such change is that the affinity of protein binding to IGF-I and the functional outcome are often dependent on posttranslational events (13) that cannot be predicted from protein abundance.

Further verification that basal IGF-I signaling was not compromised in the diabetic rat retina was provided by the finding that Akt phosphorylation was not decreased, and actually tended to be enhanced. This is noteworthy, because Akt phosphorylation is downstream not only of the IGF-IR, but also of the insulin receptor (9,32), and the diabetic rats in this study had systemic signs of chronic insulin deficiency and received no insulin replacement for at least 48 h before they were killed. This raises the possibility that Akt regulation in the retina is not responsive to circulating insulin levels, or, as proposed earlier for IGF-I, that the degree of ligand deficiency is not sufficient to compromise signaling or is compensated for by other local influences. A modest but significant increase over control was found in the levels of Akt phosphorylation in the retina of galactosemic rats, which also showed increased frequency of capillary cell apoptosis. Akt phosphorylation may reflect in these retinas the contribution of cellular stress, which can induce Akt activation through the p38/HOG1 kinase cascade (32,33). The earlier onset of retinal capillary cell apoptosis in the galactosemic than diabetic rats is in accord with our previous findings that rats with 6-month duration of galactosemia showed a significant number of acellular capillaries not noted in rats with almost 8 months of diabetes (1). The different time of onset of vascular apoptosis in diabetic and galactosemic rats, coupled with the selective effect of aminoguanidine on diabetes-induced apoptosis and retinopathy (34), indicates that different mechanisms initiate the pathway to retinal microangiopathy in the two models. Apoptosis of retinal neurons was not examined in this study, but the timing of occurrence reported by Barber et al. (4)—as early as 1 month after induction of streptozotocin diabetes in the rat—militates against a contribution from decreased IGF-I synthesis, which was as yet unaltered after 2 months of experimental diabetes.

Although a multitude of studies have investigated IGF-I in diabetic tissues and found decreased synthesis, signaling and biological consequences have largely been surmised. This work shows that reduced IGF-I mRNA levels in diabetes do not necessarily translate into decreased biological actions, at least for some time, and perhaps on account of modulatory/compensatory influences. However, IGF-I signaling could only be studied in the rat retina, and diabetic rats had shown a lesser degree of suppression of retinal IGF-I synthesis than the human diabetic donors. Hence, our data cannot exclude consequences of the more severe changes observed in human diabetes, nor can they exclude that treatment with exogenous IGF-I may have beneficial effects on the course of diabetic retinopathy. However, the data warn that exogenous IGF-I would represent a pharmacological intervention rather than necessary replacement until proof is obtained that decreased retinal IGF-I synthesis in diabetes is a primary defect (and not, for example, the result of processes that modulate IGF-I affinity for its receptor), and plays a pathogenic role in the development of retinopathy.

This work was supported by Public Health Service Grant EY09122, a grant from the Massachusetts Lions Eye Research Fund, and the George and Frances Levin Endowment. G.R. was the recipient of a fellowship from the board of trustees of the Schepens Eye Research Institute.

We are indebted to Dr. A. Kazlauskas for the gift of platelet-derived growth factor—treated endothelial cells and for critical reading of the manuscript.