Essential Role of Growth Hormone in Ischemia-Induced Retinal Neovascularization

See allHide authors and affiliations

Science  13 Jun 1997:
Vol. 276, Issue 5319, pp. 1706-1709
DOI: 10.1126/science.276.5319.1706


Retinal neovascularization is the major cause of untreatable blindness. The role of growth hormone (GH) in ischemia-associated retinal neovascularization was studied in transgenic mice expressing a GH antagonist gene and in normal mice given an inhibitor of GH secretion (MK678). Retinal neovascularization was inhibited in these mice in inverse proportion to serum levels of GH and a downstream effector, insulin-like growth factor–I (IGF-I). Inhibition was reversed with exogenous IGF-I administration. GH inhibition did not diminish hypoxia-stimulated retinal vascular endothelial growth factor (VEGF) or VEGF receptor expression. These data suggest that systemic inhibition of GH or IGF-I, or both, may have therapeutic potential in preventing some forms of retinopathy.

Neovascularization, the final common pathway in diabetic retinopathy, retinopathy of prematurity, and age-related macular degeneration can cause vision loss. Surgical ablative treatments are incompletely effective and destroy retinal tissue, causing partial visual field loss. Retinal neovascularization remains the most frequent cause of blindness (1,2).

A role for a pituitary-associated factor in diabetic retinopathy was hypothesized nearly 45 years ago when retinal neovascularization in a diabetic patient was found to regress after pituitary infarction (3). For a decade thereafter, many diabetic patients were treated by pituitary ablation. The reduction of retinopathy in these patients appeared to be related to postsurgical GH deficiency, although the role of other pituitary factors could not be eliminated (4). GH-deficient dwarfs with glucose intolerance (but without diabetes) have little retinopathy compared to diabetic controls (5). Many mitogenic effects of GH are mediated by IGF-I. In some studies of diabetic patients, serum IGF-I levels or vitreal IGF-I levels are associated with proliferative retinopathy (6). Clinical trials of early somatostatin (SS) analogs to treat diabetic retinopathy have been inconclusive (7).

To investigate the role of the GH–IGF-I –SS axis in ischemia-induced retinal neovascularization and its interaction with VEGF, currently viewed as a major effector of neovascularization (8-10), we studied this process in mice with experimentally altered levels of GH. Neovascularization was induced (11) in two transgenic mouse lines. The first, G119K, expresses a GH antagonist (Gly119to Lys in bovine GH), which results in a dwarf phenotype. The second line, E117L, expresses a GH agonist (Glu117 to Leu in bovine GH) and exhibits a giant phenotype (12). There was a 34% decrease in neovascularization (13) in GH antagonist mice at postnatal day 17 (P17) compared to nontransgenic littermates (P ≤ 0.0026) (Table 1). However, transgenic mice expressing the GH agonist E117L had no increase in retinal neovascularization compared to controls (Table 1). As assessed by histologic examination of ocular cross sections, the neovascular response of controls (Fig. 1A) was suppressed in GH antagonist transgenic mice (Fig. 1B). In controls, significant areas of neovascularization were also detected in flat-mounted whole retinas (Fig. 1C) that were perfused with a nondiffusable fluorescein-dextran solution that fills all vessels (14). Neovascularization is diminished in GH antagonist G119K retinas (Fig. 1D). No abnormal retinal or vascular development or toxicity was detectable by light microscopy.

Figure 1

Effect of GH inhibition on ischemia-induced retinal neovascularization. (A) Cross section of an eye from a nontransgenic littermate mouse, showing retinal neovascularization internal to the inner limiting membrane (arrows). (B) Cross section of an eye from a GH antagonist G119K transgenic mouse. No vascular cell nuclei are apparent internal to the inner limiting membrane. (C) Nontransgenic flat-mounted whole retina, showing extensive areas of retinal neovascularization (14) (bright fluorescence, indicated in part with arrows) that is significantly reduced in the retinas from the GH antagonist G119K transgenic mice (D).

Table 1

Inhibition of GH in vivo inhibits ischemia-induced retinal neovascularization in mice. Retinal neovascularization was induced in nontransgenic littermate mice (A1) and GH antagonist G119K transgenic mice (A2) (11). Eye sections were evaluated for neovascularization (13). GH agonist E117L transgenic mice (B4) and nontransgenic littermate controls (B3) were evaluated similarly. C57BL/6 mice (C1 to C7) were evaluated similarly after MK678 treatment (16). NS, not significant; ND, not determined.

View this table:

We also investigated the effect of MK678, an SS analog that potently inhibits GH release, on retinal neovascularization in nontransgenic mice. MK678 is long-acting and is specific for the SS subtype 2 receptor (15). Retinal neovascularization was reduced up to 44% in mice systemically treated with MK678 compared to untreated controls (16). The extent of inhibition depended on the MK678 dose (Table 1). In both the G119K transgenic and MK678-treated mice (Table 1), the extent of neovascularization correlated inversely with GH and IGF-I levels (17).

The mean serum IGF-I level was 33% less in GH antagonist mice than in controls (P ≤ 0.0001) and 83% greater in GH agonist mice than in controls (P ≤ 0.027). Mean serum GH levels from MK678-treated mice declined in proportion to the dose of MK678 (Table 1). IGF-I levels in mice treated with MK678 correlated strongly (18) with neovascularization as determined by simple linear regression analysis (Fig. 2).

Figure 2

Linear association between serum IGF-I and retinal neovascularization in mice treated with MK678 (18).y = 0.137x + 2.32 [y = average number of vascular nuclei per section per mouse, x = serum IGF-I level (ng/ml)],r 2 = 0.877, SEE = 4.76, P< 0.001.

The role of GH in hypoxia-induced retinal neovascularization was also examined directly by treating MK678-treated mice with exogenous GH. Mice were coinjected subcutaneously with MK678 and murine recombinant GH (mrGH) three times a day to approximate the normal pulsatile release of GH. Even with incomplete restoration of GH, the average number of neovascular nuclei per section per eye was increased from 56% of saline control to 64% of control (P ≤ 0.04), suggesting that MK678 suppression of retinal neovascularization was mediated at least in part through GH suppression. In parallel studies, retinal neovascularization was induced in mice treated with both MK678 and human recombinant IGF-I (hrIGF-1) (Table 1). Replacement of serum IGF-I completely restored neovascularization to control levels in MK678-treated mice, suggesting that the GH effect was mediated in large part through IGF-I. Exogenous IGF-I alone did not increase neovascularization over control levels (Table 1).

The levels of VEGF, a major stimulus for retinal neovascularization, increase in mouse retina 6 to 12 hours after the onset of hypoxia and remain elevated during the induction of neovascularization. To examine whether the GH/IGF-I effect on neovascularization was mediated directly through VEGF, we evaluated P14 (near maximum VEGF mRNA response in this model) and P17 (maximum neovascularization) retinas from MK678-treated or control mice by Northern blot (19) forVEGF and for VEGF receptor (Flk-1) transcripts. In mice with hypoxia-induced retinal neovascularization, neither MK678 treatment nor GH antagonist transgene expression (20) inhibited VEGF or Flk mRNA levels compared to controls at P14 or P17 (Fig. 3C). Immunoblot analysis (21) showed no difference in VEGF levels at P17 (Fig. 3A) or P14 (20) between transgenic GH antagonist mice (G119K) or controls, nor between MK678-treated and untreated mice at P17 (Fig. 3B) or P14 (20). These results suggest that the hypoxia response of VEGF was intact in all models tested.

Figure 3

Immunoblot and Northern analysis of VEGF in retinas from GH-inhibited mice with induced retinal neovascularization. Immunoblot analysis of total protein from retinas of (A) room-air control mice (lane 1), neovascularization-induced control mice (lane 2), and neovascularization-induced GH antagonist (G119K) mice (lane 3). (B) Neovascularization-induced untreated control mice (lane 1) and neovascularization-induced MK678-treated mice (lane 2). Lane 4, VEGF (23 kD) molecular size standard (2 ng). (A) The relative intensity of lanes normalized to lane 2 was 59% (lane 1), 100% (lane 2), and 96% (lane 3). (B) The relative intensity of lanes normalized to lane 1 was 100% (lane 1) and 105% (lane 2) (19). (C) Northern blot of total mRNA in GH-inhibited and control retinas, probed for transcripts of VEGF [3700 base pairs (bp)],Flk (6100 bp), and 36B4ribozome-associated control (1200 bp). Retinas (at least 22 per lane) were pooled for extraction of total RNA from room-air–exposed normals at P14 (lane 1), neovascularization-induced controls at P14 (lane 2) and P17 (lane 4), and neovascularization-induced MK678-treated mice at P14 (lane 3) and P17 (lane 5). Arrows indicate the ribosomal 28S and 18S mRNA markers. The relative intensities of Flk to 36B4 were 0.18 (lane 1), 0.21 (lane 2), 0.23 (lane 3), 0.19 (lane 4), and 0.30 (lane 5). The relative intensities of VEGF to36B4 were 0.06 (lane 1), 0.13 (lane 2), 0.15 (lane 3), 0.15 (lane 4), and 0.16 (lane 5) (21).

Our results indicate that inhibition of GH can inhibit ischemia-induced retinal neovascularization in vivo. Neovascularization in animals with inhibited GH secretion was completely restored with exogenous IGF-I. This is consistent with the hypothesis that the inhibition of retinal neovascularization by SS agonists is mediated by a direct lowering of GH levels resulting in a subsequent decrease in IGF-I synthesis. However, we cannot rule out other mechanisms, such as direct and indirect effects of GH and SS analogs specific to receptor 2 (22). Although VEGF is an important hypoxia-induced mediator for retinal neovascularization, our studies indicate that inhibition of GH secretion or action did not reduce the hypoxia-induced VEGF mRNA, or protein levels in vivo. Thus, IGF-I and VEGF may have distinct functions in the control of angiogenesis such as acute oxygen regulation (VEGF) versus control of neovascularization on the basis of availability of nutrients for cell division (IGF-I).

No significant increase in neovascularization was observed with the increased GH levels in giant E117L transgenic mice. This observation is in agreement with clinical studies. Patients with overexpression of GH (agromegaly) with or without concomitant diabetes do not have an increased incidence of retinopathy (23). There is a linear correlation between serum IGF-I levels and retinal neovascularization in GH-inhibited mice at low and normal serum IGF-I levels (Fig. 2). However, addition of exogenous IGF-I to augment serum levels did not increase retinal neovascularization. These data suggest a permissive role in retinal neovascularization for both GH and IGF-I, with a plateau in response at higher doses. IGF-I receptors are distributed widely in the eye and, although present in vascular cells, they predominate in the neural retina (24). Thus, the mechanism of action of IGF-I on the development of retinal neovascularization may be complex and indirect.

In our model of aggressive retinopathy, neovascularization was only partially suppressed by inhibition of GH. Insufficient reduction of GH and serum IGF-I levels or GH-independent local IGF-I production may account for the partial inhibition, as could the possibility that the GH–IGF-I axis is only one of multiple control mechanisms regulating neovascularization. However, the 30 to 44% inhibition of neovascularization that we observed in mice treated with MK678 would be clinically significant. The therapeutic effect is comparable to that observed in the clinical trials that evaluated laser scatter photocoagulation and cryotherapy for diabetic retinopathy and retinopathy of prematurity, respectively (25). In this model inhibition of αv integrins (downstream of VEGF and basic fibroblast growth factor and perhaps other effectors) greatly reduces retinal neovascularization (26). Inhibition of either VEGF or GH results in comparable reductions in retinal neovascularization. Inhibition of any one pathway may preferentially affect neovascularization in different patholologic (retinopathy, tumor growth) and beneficial (wound healing, repair of ischemic myocardium, ovulation and fetal development) processes. Isolated GH deficiency in humans appears to allow ovaluation and fetal development (27). Thus inhibitors of GH and IGF-I action may be useful for inhibition of neovascularization in specific vascular areas alone or when given in conjunction with other angiogenic inhibitors.

  • * To whom correspondence should be addressed. E-mail: smith_lo{at}


View Abstract

Stay Connected to Science

Navigate This Article