Vascular endothelial growth factor in the ischemic retina and its regulation by somatostatin


Address correspondence and reprint requests to Giovanni Casini, PhD, Dept. for Innovation in Biological, Agro-food and Forest systems, University of Tuscia, Largo dell’Università snc, blocco D, 01100 Viterbo, Italy. Email:


J. Neurochem. (2012) 120, 818–829.


In a retinal ischemic ex vivo model, we have reported protective effects of somatostatin (SRIF) receptor 2 (sst2). As an ischemic condition not only causes cell death but also induces a vascular response, we asked whether vascular endothelial growth factor (VEGF) is altered in this model and whether its expression, release or localization are affected by sst2 activation. Ex vivo retinas of wild-type (WT) and sst1 KO mice (which over-express sst2) were incubated in ischemic conditions with SRIF, octreotide (OCT) or a VEGF trap. Ischemia in WT retinas caused increase of VEGF release and decrease of VEGF mRNA. Both effects were counteracted by SRIF or OCT. VEGF immunoreactivity was in retinal neurons and scarcely in vessels. Ischemia caused a significant shift of VEGF immunoreactivity from neurons to vessels. The increase of vascular VEGF was reduced in sst1 KO retinas and in WT retinas treated with SRIF or OCT. VEGF trap also limited this increase, demonstrating that vascular VEGF was of extracellular origin. Together, the data show a VEGF response to ischemia, in which VEGF released by damaged neurons reaches the retinal capillaries. The activation of sst2 protects neurons from ischemic damage, thereby limiting VEGF release and the VEGF response.

Abbreviations used:

calcium-binding protein 5


inner nuclear layer




neurokinin 3




somatotropin release-inhibiting factor


vascular endothelial growth factor



In mammalian retinas, vascular endothelial growth factor (VEGF) promotes both normal vascular development (Stone et al. 1995) and pathological angiogenesis (Yla-Herttuala et al. 2007; Zhang et al. 2009). Different retinal diseases, such as diabetic retinopathy, macular degeneration, and retinopathy of prematurity, are characterized by aberrant neovascularization and increased vascular permeability caused primarily by VEGF (Rajappa et al. 2010), and current anti-VEGF drugs have shown marked beneficial effects in proliferative retinal diseases (Abdallah and Fawzi 2009; Chiang and Regillo 2011; Goyal et al. 2011). A variety of experimental data from in vivo models have demonstrated that hypoxic conditions induce a significant VEGF up-regulation and the growth of new, aberrant vessels. Both these effects are counteracted by the activation of the somatostatin (or somatotropin release-inhibiting factor, SRIF) subtype 2 receptor (sst2) obtained using either retinas of sst1 knock-out (KO) mice, which are characterized by over-expression and constitutive over-activation of sst2 (Dal Monte et al. 2003; Bigiani et al. 2004; Pavan et al. 2004; Casini et al. 2005), or the use of the sst2-preferring agonist octreotide (OCT) (Dal Monte et al. 2007, 2009).

We have recently developed and used an ischemic ex vivo model, in which mouse retinas are incubated for one hour in conditions of metabolic inhibition and anoxia (Catalani et al. 2007; Cervia et al. 2008). The advantage of ex vivo models is the option to make direct observations and measurements of cellular responses to chemicals in a defined extracellular environment (Mastrodimou et al. 2005, 2008; Massote et al. 2008). In addition, blood flow effects are also excluded. Our model is characterized by marked apoptotic cell death, increased glutamate release and pathological changes in selected retinal cell populations as, for instance, the rod bipolar cells (Catalani et al. 2007). As an ischemic condition not only causes cell death but also induces a vascular response, we asked whether VEGF is also affected in this model. It should be noted that ischemia is different from hypoxia. Hypoxia refers to a reduction of either oxygen supply or utilization, while ischemia describes a reduction in blood supply leading not only to decreased oxygen delivery but also to decreased glucose supply and limited or no removal of damaging cellular metabolites. Ischemia always has a component of hypoxia/anoxia, but hypoxia/anoxia does not imply ischemia. Retinal ischemia is a common clinical entity and, due to relatively ineffective treatment, remains a common cause of visual impairment and blindness (Osborne et al. 2004; Stitt et al. 2011).

Using sst1 KO mice or SRIF analog administrations, we have demonstrated that activation of sst2 in the ex vivo ischemic retina protects from the damage caused by ischemia (Catalani et al. 2007; Cervia et al. 2008). We have not yet investigated, in this model, the existence of possible effects on VEGF caused by sst2 activation. Therefore, the aims of the present paper were: (i) to identify the retinal elements containing VEGF, confirming and expanding our previous observations (Dal Monte et al. 2007); (ii) to verify changes in VEGF expression, release or distribution patterns in the retina after the induction of ischemia; (iii) to elucidate possible effects of sst2 activation on the VEGF response using sst1 KO mice or administration of SRIF or OCT.

Materials and methods

Drugs and chemicals

A rabbit antiserum directed to VEGF (catalog number: sc-507) and horseradish peroxidase-linked secondary antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). A polyclonal rat VEGF antibody (catalog no.: 103-M57) and the recombinant murine VEGF164 were from Reliatech GmbH (Braunschweigh, Germany). Rat CD31 antibody was purchased from BD Pharmingen (San Diego, CA, USA). The sources of other primary antibodies are specified in Table 1. Alexa Fluor 546 and Alexa Fluor 488 were from Molecular Probes (Eugene, OR, USA). SRIF (SRIF-14) was purchased from Bachem (Bubendorf, Switzerland), and OCT was a gift of Prof. D. Hoyer and Dr H. Schmid (Novartis Pharma, Basel, Switzerland). The recombinant mouse VEGF R1/Flt-1 FC chimera (VEGF trap, catalog number 321-FL) was purchased from R&D Systems (Minneapolis, MN, USA). Where not specified, chemicals and reagents were purchased from Sigma-Aldrich (St Louis, MO, USA).

Table 1.   Antibodies used in double label experiments to evaluate the localization patterns of VEGF in retinal neurons
AntigenHostSourceDilutionLabeled cells (Reference)
Cone arrestinRabbitDr C.M. Craft1 : 500Cone photoreceptors (Zhu et al. 2002)
MAb115A10MouseDr S.C. Fujita1 : 3000Rod bipolar cells, ON-type cone bipolar cells (Greferath et al. 1990)
CaBP5RabbitDr F. Haeseleer1 : 500Rod bipolar cells, a population of ON- and a population of OFF-type cone bipolar cells (Haverkamp et al. 2003)
NK3RabbitDr N.W. Bunnett1 : 400Two populations of OFF-type cone bipolar cells (Haverkamp et al. 2003)
BassoonMouseStressGen biotech1 : 500Synaptic ribbons associated with ON-type contacts in the OPL (Brandstatter et al. 1999)
VimentinMouseSigma1 : 50Müller cells (Drager et al. 1984)
β-Tubulin IIIMouseSigma1 : 400Ganglion cells (Watanabe et al. 1991; Snow and Robson 1994)


Experiments were performed on mouse retinas of wild-type (WT, C57BL/6) and sst1 KO (Kreienkamp et al. 1999) strains of both sexes at 6–8 weeks after birth (20–30 g body weight). Animals were kept in a regulated environment (23 ± 1°C, 50 ± 5% humidity) with a 12-h light/dark cycle (lights on at 08:00 am) with food and water ad lib. In all experiments, the mice were anesthetized by i.p. injection of Avertin (1.2% tribromoethanol and 2.4% amylene hydrate, 0.02 mL/g body weight) and killed by cervical dislocation. The experiments were performed in compliance with the Italian law on animal care N° 116/1992 and the EEC/609/86. The experimental protocols were also approved by the Bioetic Committee of the University of Tuscia. All efforts were made to reduce both animal suffering and the number of animals used.

Western blot

Western blot was performed following published protocols (Dal Monte et al. 2009), as detailed in Appendix S1, to confirm VEGF antibody specificity.

Ex vivo ischemic treatment and drug administration

Isolated retinas were incubated in the ischemic solution for 1 h, as previously described (Catalani et al. 2007; Cervia et al. 2008). The native peptide SRIF or the sst2-preferring agonist OCT were applied at the beginning of the incubation period at 1 μM, a concentration giving maximal receptor occupancy in different systems (Weckbecker et al. 2003), including mouse and rat retina (Bigiani et al. 2004; Pavan et al. 2004; Mastrodimou et al. 2005, 2006a,b, 2008). The VEGF trap was also applied for 1 h at 1 μM. It is a recombinant soluble Flt-1/Fc chimera, which binds with high affinity to all forms of VEGF and has favorable kinetics that provide for long-term VEGF blockade. All experiments were performed at the same time of the day (between 09:00 and 11:00 am) to exclude possible circadian influences. The data were collected from both male and female WT and sst1 KO mice; the results were combined as there were no apparent gender differences.

Enzyme-linked immunosorbent assay

VEGF was detected and quantified in the supernatant of WT retinas (VEGF release) by means of a commercially available enzyme-linked immunosorbent assay (ELISA) kit (Quantikine Mouse VEGF ELISA kit; R&D Systems), according to the manufacturer’s protocol. Data from three independent experiments were collected as picograms VEGF per mL surnatant and, after statistics (raw data), averaged in the same graph.

Real-time RT-PCR (QPCR)

QPCR experiments were carried out on ischemic and non-ischemic WT retinas as previously described (Cervia et al. 2008) and detailed in Appendix S1.

Immunohistochemical procedures

Immunohistochemistry was performed following published protocols (Casini et al. 2004; Catalani et al. 2009) using rabbit or rat anti-VEGF (1 : 200) and appropriate secondary antibodies conjugated with Alexa Fluor fluorescent dyes (Alexa Fluor 546 or Alexa Fluor 488). Antibody specificity was assessed with western blots (see above) or by blocking the VEGF antibodies with 10 μM of recombinant murine VEGF164.

To evaluate the presence of VEGF in endothelial cells, the VEGF antibody raised in rabbit was used in conjunction with an antibody raised in rat and directed to CD31 (1 : 100 or 1 : 50 dilution), which stains blood vessel endothelial cells (Favaloro et al. 1989). Freshly dissected WT retinas were used to elucidate the VEGF immunostaining patterns in the mouse retina, while retinas incubated in ischemic or in control conditions were used to examine changes in this pattern caused by ischemia. VEGF patterns in freshly dissected retinas were also investigated with double-label immunohistochemistry using VEGF antibodies in conjunction with each of the antibodies listed in Table 1, following published protocols (Catalani et al. 2009). Details on double-label protocols and image acquisition/optimization are available in Appendix S1.

Quantitative analysis of VEGF immunoreactivity in retinal vessels

This analysis was performed on VEGF/CD31 double-labeled cryostat sections from retinas incubated for 1 h in ischemic or in control conditions. We could not use retinal whole mounts because the ischemic treatment markedly affected antibody penetration in whole retinas. Preparation of retinal sections and collection of slides was as previously described (Catalani et al. 2007; Cervia et al. 2008). Briefly, 10 μm cryostat retinal sections were used. All retinas were cut with the same temporal-to-nasal orientation. Consecutive sections were alternately put onto a series of five slides, so that on each slide the sections were spaced every 50 μm. Five sections were put on each slide, and five series of slides were prepared from each retina. Corresponding series of sections, from control, ischemic and drug treated retinas, were used in each experiment. Three retinas were analyzed for each experimental condition, and at least two series of slides from each retina were used for the quantitative analysis.

The CD31 and VEGF immunostainings were evaluated to obtain a measure of the presence of VEGF in retinal endothelial cells. The measurements were obtained with a system of computer-assisted image analysis using Zeiss Axiovision 4 and KS-300 software (Carl Zeiss). The CD31 and the VEGF images relative to the same field were treated as summarized in Fig. 1. Briefly, the CD31 image was thresholded (white staining over black background), while the VEGF image was turned into grayscale and normalized to the background. The CD31, thresholded image and the VEGF, normalized image were superimposed in two different layers. Finally, the CD31 positive area was selected and the mean value of the pixels of the VEGF image was calculated within the CD31 selection. This method provided a measure of the VEGF immunolabel intensity within the area of the blood vessels.

Figure 1.

 Schematic of the image analysis method used to quantify VEGF immunolabeling in retinal endothelial cells. See the text for further details. Scale bar: 20 μm.


Upon verification of normal distribution, statistical significance of raw data between the groups in each experiment was evaluated using anova followed by multiple comparison Tukey’s Multiple Comparison post-test. The GraphPad Prism software package (Graph Software, San Diego, CA, USA) was used. The results were expressed as mean ± SEM.


VEGF immunoreactivity in the mouse retina

As shown in the western blots of Fig. 2 (Fig. 2b represents the western blot of a retinal sample of higher total protein concentration than in Fig. 2a), the VEGF polyclonal antibody from Santa Cruz darkly labeled a protein band in cytosolic extracts of freshly dissected WT mouse retinas, corresponding to the expected molecular weight of VEGF (20 kDa). In addition, VEGF immunoreactivity was also in a lightly stained band around 40 kDa, which is likely to represent VEGF in the form of dimers. This interpretation is supported by similar observations in the rat myocardium, where VEGF has been observed to migrate in sodium dodecyl sulfate–polyacrylamide gel electrophoresis as a monomer of 20 kDa and a dimer of approximately 40 kDa (Ray et al. 2000). Increasing the amount of total protein resulted in darker immunolabeling of protein bands. Similar results were obtained with the VEGF antiserum from Reliatech (data not shown).

Figure 2.

 VEGF immunoreactivity as evaluated by western blot using the rabbit polyclonal antibody. Two different amounts of total proteins were separated by 10% SDS–PAGE (lane a, 30 μg; lane b, 60 μg). The VEGF antibody revealed two bands corresponding to the monomer (20 kDa) and the dimer (40 kDa) forms of VEGF. The bands corresponding to 60 μg of total protein were darker than those corresponding to 30 μg of total protein. β-actin was used as the loading control.

The localization patterns of VEGF were studied by immunohistochemistry in sections of freshly dissected WT retinas. Both VEGF antibodies gave comparable results, and both were blocked by pre-incubation with recombinant murine VEGF164 (Figure S1). The following data are mainly based on retinal samples treated with the Santa Cruz antiserum, which provided the best signal-to-noise ratio. As shown in Fig. 3a, VEGF immunostaining was detected in all retinal layers. In confocal images scanned through whole retinas, VEGF immunostaining could be detected in the outer nuclear layer in cone photoreceptor somata (Fig. 3b) and axons (Fig. 3c). In the outer plexiform layer (OPL, Fig. 3d), VEGF immunolabeling was uniformly diffused, although intensely labeled puncta were also detected. Several bipolar cell somata displaying VEGF immunostaining were visible in the mid-distal inner nuclear layer (INL; Fig. 3e), and their axons could be followed through the proximal INL (Fig. 3f) until they dispersed into the inner plexiform layer, where a diffuse, grainy VEGF immunostaining was evenly distributed (Fig. 3g). A similar immunostaining pattern was in the ganglion cell layer, where VEGF immunoreactive somata could also be seen together with numerous unlabeled somata (Fig. 3h). Only rarely was VEGF immunoreactivity detected in profiles resembling retinal blood capillaries. Figure 3i documents the identical VEGF immunostaining pattern obtained using the polyclonal antibody from Reliatech.

Figure 3.

 Patterns of VEGF expression in the mouse retina. (a) Confocal image of a cryostat section stained with the Santa Cruz VEGF polyclonal antibody. Immunolabeling is distributed to all retinal layers. (b–h) Confocal images taken at different depths of whole mount preparations. (b) VEGF immunoreactive cone photoreceptor somata in the outer nuclear layer (ONL) whose axons are shown in panel (c). (d) VEGF immunostaining in the outer plexiform layer (OPL). (e) VEGF immunolabeled bipolar cell somata in the mid-distal inner nuclear layer (INL) whose axons are shown in panel (f) (proximal INL). (g) VEGF immunostaining in the inner plexiform layer (IPL). (h) VEGF immunostaining in the ganglion cell layer (GCL). The arrow in panel (a) points to a VEGF immunoreactive, large cell body in the GCL, likely belonging to a ganglion cell, while the arrowheads in H indicate unlabeled somata surrounded by VEGF immunostaining in the GCL and likely representing displaced amacrine cells. Scale bar: 10 μm. The inset (i) shows a retinal section with a similar VEGF immunostaining pattern obtained with the VEGF antibody purchased from Reliatech (confocal image). Scale bar: 20 μm. POS, photoreceptor outer segments.

Double-label experiments were performed to analyze VEGF localization in cone photoreceptors, bipolar cells, ganglion cells and to test VEGF presence in Müller cells. VEGF was expressed in the totality of cone photoreceptors labeled with cone arrestin antibodies (Figure S2). Regarding VEGF localization to bipolar cells, it was not in MAb115A10 immunostained bipolar cells (Fig. 4a), which include the rod bipolar cells and the ON-type cone bipolar cells. Consistently, no associations were detected in the OPL between VEGF immunostained profiles and bassoon (which labels ribbons at photoreceptor/ON-type cell synapses, Fig. 4b). In contrast, many VEGF immunoreactive bipolar cells were also neurokinin 3 (NK3, Fig. 4c) or calcium-binding protein 5 (CaBP5, Fig. 4d) immunoreactive. These colocalization patterns indicate that VEGF is not localized to ON-type bipolar cells, but it is in one or more populations of OFF-type cone bipolar cells. The presence of VEGF in all ganglion cells was confirmed by double labels with β-Tubulin III antibodies (Figure S3a). Finally, lack of colocalization of VEGF and vimentin immunostainings indicated that VEGF is not in Müller cells (Figure S3b).

Figure 4.

 Confocal images from cryostat sections double labeled with VEGF and MAb115A10 (a), bassoon (b), NK3 (c), or CaBP5 (d). The arrows in panels (c) and (d) indicate VEGF/NK3 and VEGF/CaBP5 double-labeled cell bodies, respectively. The arrowheads point to bipolar cell somata single labeled with VEGF. Abbreviations as in Fig. 3. Scale bars: 20 μm in panels (a, c and d); 10 μm in (b); 5 μm in the inset.

To better elucidate the presence of VEGF in retinal blood capillaries, double-label experiments were performed in whole retinas using VEGF antibodies in conjunction with antibodies directed to CD31. The analysis showed only faint VEGF immunostaining associated with retinal blood vessels (Fig. 5).

Figure 5.

 Confocal images from a whole mount preparation of a WT, untreated retina scanned at the level of the border between INL and IPL showing retinal blood capillaries double labeled with VEGF and CD31. The boxed areas in the upper row are shown at higher magnification in the lower row. Scale bars: 40 and 20 μm, respectively.

VEGF release and VEGF mRNA expression in ischemia

The amount of VEGF released after 1 h incubation by control and ischemic retinas, in the absence or in the presence of SRIF or OCT, was measured in the incubation medium using ELISA. As summarized in the graph of Fig. 6a, the ischemic treatment caused a dramatic increase of VEGF release (from 3.4 ± 1.0 to 42.8 ± 4.7 pg/mL). SRIF or OCT did not affect VEGF release in control conditions, but they significantly decreased, by about 45%, the release of VEGF from ischemic retinas (from 42.8 ± 4.7 to 22.7 ± 1.0 or 23.4 ± 0.7 pg/mL, respectively).

Figure 6.

 (a) VEGF release measured with ELISA in the surnatant of WT retinas in different experimental conditions. (b) VEGF mRNA measured with QPCR in WT retinas in different experimental conditions. *Significantly different from control; §significantly different from ischemic. Two or three symbols indicate, respectively, p < 0.01 or p < 0.001.

VEGF mRNA expression was evaluated in WT retinas in various experimental conditions using QPCR. As shown in Fig. 6b, the incubation of retinas in the ischemic solution caused a marked reduction (about 78%) of VEGF mRNA levels. SRIF or OCT did not affect VEGF mRNA levels in control conditions. However, in the presence of SRIF or OCT, VEGF mRNA levels in ischemic retinas recovered to values that were not significantly different from those in control retinas.

VEGF immunostaining in blood vessels of ischemic mouse retinas

In addition to pharmacological treatments with SRIF or OCT, in these experiments we also used retinas of sst1 KO mice, in which sst2 is over-expressed and constitutively activated (Dal Monte et al. 2003; Bigiani et al. 2004; Pavan et al. 2004; Casini et al. 2005) and which have been demonstrated to be protected, to a certain extent, from ischemic damage (Catalani et al. 2007).

As reported above, VEGF localization in endothelial cells identified by CD31 immunostaining is scarce in normal mouse retinas. Consistently, we found only little VEGF immunolabeling in the vessels of retinas incubated in control conditions both in WT (Fig. 7a) and sst1 KO (data not shown), and the quantitative measurements did not reveal statistical differences (Fig. 8). In these retinas, VEGF immunoreactivity was mainly seen in neuronal elements, following the same pattern as that seen in freshly dissected WT retinas (see Fig. 3). Both in WT and in sst1 KO retinas, the ischemic treatment determined the disappearance of VEGF immunostaining from retinal neuronal cells, while VEGF immunoreactivity showed increased overlapping with CD31 immunostaining (Fig. 7b and c). In particular, VEGF immunoreactivity almost doubled in the blood vessels of both ischemic WT and ischemic sst1 KO retinas; however, it is important to note that the levels in sst1 KO were significantly less pronounced than in WT retinas (Fig. 8).

Figure 7.

 Retinal sections from WT control (a), WT ischemic (b) or sst1 KO ischemic (c) mice showing representative patterns of VEGF and CD31 double labeling in retinal vessels. The patterns in sst1 KO control retinas (data not shown) were similar to those of WT controls depicted in A. VEGF immunolabeled cells in the INL can be observed in control retinas but not in ischemic retinas, either WT or sst1 KO. Abbreviations as in Fig. 3. Scale bar: 20 μm.

Figure 8.

 Quantitative analysis of VEGF and CD31 double labeling in retinal vessels. The values are normalized to those of WT control. *Significantly different from WT control; §significantly different from WT ischemic; #significantly different from sst1 KO control; +significantly different from sst1 KO ischemic. One, two or three symbols indicate, respectively, p < 0.05, p < 0.01 or p < 0.001.

As depicted in the histograms of Fig. 8 (see also Figure S4), the treatment of WT ischemic retinas with SRIF or OCT resulted in a decrease of VEGF immunolabeling of retinal vessels by about 63% and 87%, respectively, that is to levels that were similar to those measured in ischemic sst1 KO retinas. In contrast, in the ischemic sst1 KO retinas the treatment with SRIF or OCT (Figure S5) did not produce the effects that could have been expected. Indeed, instead of further reducing VEGF immunostaining in the blood vessels, the treatment with SRIF produced a trend towards an increase, although not significant, while after treatment with OCT the overlapping of VEGF with CD31 increased by about 32% (Fig. 8).

The effect of blocking extracellular VEGF

Our results are consistent with the possibility that VEGF, massively released by damaged neurons, reaches the retinal capillaries where it is likely to initiate the mechanisms for the formation of new vessels. To test this hypothesis and to exclude de novo synthesis of VEGF by endothelial cells of retinal blood vessels, we used a VEGF trap to block extracellular VEGF and prevent possible interactions of extracellular VEGF with retinal vessels. As shown in Fig. 8, in ischemic retinas treated with VEGF trap the amount of VEGF immunoreactivity associated with blood capillaries was significantly reduced and was not significantly different from that measured in control retinas.


The present work shows that retinal VEGF patterns are profoundly affected by the onset of an ischemic state. These changes may represent the fast response of the VEGF system to severe shortage of nutrients and oxygen in retinal neurons. In addition, our data show that activation of sst2, which is known to protect the retina from ischemic damage, also reduces the VEGF response to ischemia.

VEGF immunostaining patterns in normal mouse retinas

Our immunohistochemical data show that VEGF is mainly localized to retinal neurons and scarcely to retinal vessels, confirming and expanding our previous observations (Dal Monte et al. 2007). The specificity of VEGF immunostaining was validated by western blot and by blocking the primary antibody with its antigen. In addition, two different VEGF antibodies, directed to different epitopes, produced the same immunostaining patterns. The double-label data demonstrate that VEGF is in cone photoreceptors, in OFF-type cone bipolar cells, and in ganglion cells, and they exclude the presence of VEGF in Müller cells. In particular, VEGF immunoreactivity was observed in OFF-type cone bipolar cells expressing NK3 or CaBP5. As it has been demonstrated that NK3 and CaBP5 immunoreactivities do not coexist in the same bipolar cells of the mouse retina (Haverkamp et al. 2003), these observations indicate that VEGF is localized to at least two different populations of OFF-type cone bipolar cells.

In rodent retinas, VEGF immunoreactivity and/or mRNA have been observed mainly in the ganglion cell layer and INL (Pierce et al. 1995; Dorey et al. 1996; Yi et al. 1998; Yamada et al. 1999; Kim et al. 2003; Gariano et al. 2006; Kilic et al. 2006; Lee et al. 2007; Prokosch et al. 2011). These data confirm our finding of VEGF immunoreactivity in ganglion cells and are consistent with the presence of VEGF in bipolar cells in the INL. VEGF expression has also been reported in Müller cells of mouse and rat retinas (Pierce et al. 1995; Dorey et al. 1996; Yi et al. 1998). This seems to be in contrast with our findings, however it must be noted that VEGF in Müller cells of mouse retinas has been documented at postnatal day 17, and it becomes evident only after hypoxic treatment (Pierce et al. 1995). Finally, the reported presence of VEGF mRNA in the outer nuclear layer of mouse retinas (Kilic et al. 2006) is consistent with our observation of VEGF immunoreactivity in cone photoreceptors.

The VEGF response to ischemia

As noted previously, retinal neurons are severely affected by an ex vivo ischemic treatment (Mastrodimou et al. 2005, 2008; Catalani et al. 2007; Cervia et al. 2008) and large amounts of neurons undergo apoptotic cell death (Catalani et al. 2007). In such conditions, mRNA transcription is likely to be down-regulated, and perhaps only the expression of apoptotic genes is maintained or increased in the early phases of ischemia, as demonstrated for caspase 3 mRNA (Catalani et al. 2007). The drastic decrease of VEGF mRNA observed in the ischemic retina is likely to reflect this condition. Consistently, we found that VEGF immunoreactivity in neurons is also reduced to undetectable levels during ischemia. However, the concomitant massive increase of extracellular VEGF measured with ELISA strongly suggests that neurons, in addition to drastically reducing VEGF mRNA expression, release all of their VEGF content in the extracellular space at the onset of ischemia.

Our data also show that ischemia induces a significant shift of VEGF immunoreactivity from neuronal to perivascular location. This observation can be explained either with a conspicuous increase of new VEGF expression in endothelial cells or with extracellular VEGF being “taken up” by endothelial cells. To test these possibilities, we used a VEGF trap to block extracellular VEGF and demonstrated that the VEGF accumulated in endothelial cells is, in fact, of extracellular origin and not produced by endothelial cells through de novo synthesis. The extracellular VEGF may enter endothelial cells through receptor-mediated internalization; however, the elucidation of this mechanism will require further investigation. VEGF immunoreactivity in retinal vessels has been observed to increase in mouse models of oxygen-induced retinopathy (May et al. 2006; Dal Monte et al. 2007) or in the presence of retinal diseases characterized by neoangiogenesis (Lutty et al. 1996; Mathews et al. 1997; Cao et al. 1999; Ellis et al. 2000; Kim et al. 2003; Verma et al. 2010). In addition, internalization of VEGF mediated by VEGF receptor 2 has been reported in in vitro preparations of endothelial cells (Lampugnani et al. 2006; Santos et al. 2007; Chen et al. 2010; Zhao et al. 2010) as well as in the endothelium of tumor vessels (Falcon et al. 2011). In summary, our findings demonstrate a VEGF response to ischemia, which begins with the arrest of VEGF expression and massive VEGF release by retinal neurons. Then, the released VEGF reaches the endothelial cells of retinal vessels where it may initiate the vascular response to ischemia.

Our finding of drastically reduced VEGF mRNA in the ischemic retina is in apparent contrast with the stimulating effect of hypoxia on VEGF expression mediated by hypoxia-inducible factor (Arjamaa and Nikinmaa 2006; Campochiaro 2006), and with the reported increase of VEGF expression in hypoxic retinal models (Dal Monte et al. 2007, 2009; Kociok et al. 2007). Although both our ischemic model and the hypoxic retinas are characterized by low oxygen availability, and therefore by the likely activation of hypoxia-inducible factor, at least two important facts should be considered: i) the time of survival of the retina during the 1 h incubation in ischemic conditions may be too short to activate transcription of VEGF mRNA at detectable levels; ii) in the ischemic ex vivo model, neurons (which appear to be the main sources of VEGF in the mouse retina, as shown by our immunohistochemical data) are incapable of increasing VEGF expression because they undergo extensive apoptotic cell death (Catalani et al. 2007). In contrast, neuronal populations in hypoxic retinal models in vivo do not seem to be affected by cell death, and they may increase their VEGF expression over an extended period of time.

The effect of sst2 on the VEGF response to ischemia

Our data show that the VEGF response to ischemia is significantly contrasted by sst2 activation. Indeed, we observed attenuated VEGF response both in sst1 KO retinas, in which sst2 is over-expressed and constitutively over-active (Dal Monte et al. 2003; Bigiani et al. 2004; Pavan et al. 2004; Casini et al. 2005), and in WT retinas treated with exogenous SRIF or OCT. These results are in line with previous observations demonstrating that sst2 activation counteracts the effects of ischemia (Celiker and Ilhan 2002; Mastrodimou et al. 2005; Catalani et al. 2007; Cervia et al. 2008). In addition, the observed reduction of VEGF immunostaining in retinal vessels of WT ischemic retinas treated with SRIF or OCT and in sst1 KO ischemic retinas (when compared to WT) is consistent with the reported decrease of VEGF immunoreactivity in the blood vessels of sst1 KO hypoxic retinas compared to hypoxic WT (Dal Monte et al. 2007).

Interestingly, we observed that in ischemic sst1 KO retinas the pharmacological activation of sst2 by SRIF or OCT does not further reduce the VEGF response, but instead tends to increase it. Although these results may appear in contrast with our hypothesis, they are consistent with our previous data reporting increased cell death in such conditions (e.g. sst1 KO ischemic retinas treated with SRIF or OCT) (Cervia et al. 2008) and can be explained by assuming that over-expressed sst2 are rapidly desensitized by agonists, thus resulting in a decrease of their functional activity. Experimental data at pharmacological and molecular level have been provided to support this conclusion in our model (Cervia et al. 2008).

The reduced vascular localization of VEGF brought about by sst2 in the ischemic retina is likely to be due to the strong neuroprotective action of sst2 in this model (Mastrodimou et al. 2005; Catalani et al. 2007; Cervia et al. 2008). Indeed, if the VEGF accumulating in the vessels of the ischemic retina derives from dying neurons, a neuroprotective agent, by preventing neuronal death, should also prevent VEGF release. This interpretation is supported by our findings of increased VEGF mRNA (indicating better neuronal survival) concomitant with decreased VEGF release in the ischemic retinas in the presence of SRIF or OCT.

Recent findings indicate that sst2A immunoreactivity can be detected at the level of retinal vessels in hypoxic retinas (Dal Monte et al. 2010). Although we did not observe the presence of sst2A in capillaries of retinas subjected to ischemic treatment (unpublished), in principle we cannot exclude that the effects of sst2 activation on the VEGF response may also include some action of SRIF mediated by sst2 at the level of retinal vessels.


In conclusion, our data show that ischemia in the retina induces a VEGF response: VEGF mRNA expression is drastically reduced, extracellular VEGF is increased, and the endothelial cells of retinal vessels become extensively VEGF immunolabeled. Our interpretation of these observations is that neurons, severely damaged by ischemia, drastically reduce gene expression (including VEGF expression) and release all of their VEGF content. Then, the VEGF massively released by the dying neurons reaches the retinal capillaries where it is likely to initiate the mechanisms for the formation of new vessels. The activation of sst2 protects neurons from apoptosis, thereby limiting VEGF release and the VEGF response.


We wish to thank Dr Cheryl M. Craft, of the Mary D. Allen Laboratory in Vision research, Doheny Eye Institute, Los Angeles, CA, for cone arrestin antibody; Dr Shinobu C. Fujita, of the Mitsubishi Kagaku Institute of Life Sciences, Tokyo, Japan, for MAb115A10 antibody; Dr Francoise Haeseleer, of the University of Washington, Seattle, WA, for CaBP5 antibody; Dr Nigel W. Bunnett, of UCSF, for NK3 antibody; Dr daniel Hoyer, of Novartis Pharma, Basel, Switzerland, for OCT. We also thank Dr Cristiano Papeschi (Bioetic Committee of the University of Tuscia) for animal care and Dr Gabriella Gambellini (Interdepartmental Center of Electron Microscopy, University of Tuscia) for her help with confocal microscopy. This work was supported by fundings from the Italian Ministry of Education, University and Research (PRIN2005, grant no. 2005052312). The authors declare no conflicts of interest.