Immunocytochemical localization of vascular endothelial growth factor in neurons and glial cells of human retina

Brain Research 969 (2003) 195–204 www.elsevier.com / locate / brainres

Research report

Immunocytochemical localization of vascular endothelial growth factor in neurons and glial cells of human retina Edward V. Famiglietti a,b , *, Edward G. Stopa a , Edward D. McGookin c , Philip Song a , Victoria LeBlanc a , Barbara W. Streeten d a

Department of Pathology, Division of Neuropathology, Brown University School of Medicine, Rhode Island Hospital, APC-12, 593 Eddy Street, Providence, RI 02903, USA b Departments of Neuroscience and Ophthalmology, Brown University School of Medicine, Rhode Island Hospital, 593 Eddy Street, Providence, RI 02903, USA c Department of Pediatrics, Brown University School of Medicine, Rhode Island Hospital, 593 Eddy Street, Providence, RI 02903, USA d Departments of Ophthalmology and Pathology, State University of New York Health Sciences Center at Syracuse, Syracuse, NY, USA Accepted 7 October 2002

Abstract In order to establish the cellular and subcellular localization of the chemokine protein, vascular endothelial growth factor (VEGF) or vascular permeability factor, in adult human retina, we employed immunocytochemistry with double immunolabeling, using a primary ¨ antibody to amino acids 1–10 of VEGF, together with antibodies to vimentin (intermediate filaments, labeling Muller cells) or to ¨ neuron-specific enolase (labeling retinal neurons). In adult human retina, VEGF-like immunoreactivity (VEGF-IR) is found in Muller cell processes, where typically it is found in the cytoplasm in close association with Vimentin-labeled (VM-IR) intermediate filaments. ¨ VEGF-IR is sometimes found diffusely in Muller cell bodies and nuclei. VEGF-IR is found in all major classes of retinal neurons, as demonstrated by co-localization with neuron-specific enolase (NSE)-IR, but is especially prominent in cell bodies of amacrine cells (ACs) (including displaced ACs) and ganglion cells (GCs). Generally, VEGF-IR is more prominent in the nucleus, while NSE-IR is more prominent in the cytoplasm and neurites. In blood vessels, VEGF-IR co-localizes with VM-IR, marking blood vessel endothelial cells, whereas NSE-IR apparently marks the layer of smooth muscle cells. These cellular findings regarding the retinal localization of VEGF-IR are consistent with VEGF synthesis in and its export from retinal neurons, particularly amacrine and ganglion cells, as well as in glia, ¨ specifically Muller cells, and suggest that retinal neurons normally provide continuous trophic support for their retinal blood supply.  2002 Elsevier Science B.V. All rights reserved. Theme: Sensory systems Topic: Retina and photoreceptors ¨ Keywords: VEGF; Human retina; Retinal ganglion cells; Retinal amacrine cells; Retinal Muller cells

1. Introduction The cytokine, vascular endothelial growth factor (VEGF), is a 46 kDa protein, originally isolated from pituitary gland [12,19], which is synthesized and secreted by many different cell types. It exists in multiple homodimeric forms, yielded by RNA splicing. In human tissue, *Corresponding author. Tel.: 11-401-751-6316; fax: 11-401-4448514. E-mail address: edward [email protected] (E.V. Famiglietti). ]

these consist of a diffusible, 121 amino acid form, VEGF 121 and three heparin-binding forms: VEGF 165 , VEGF 189 , and VEGF 206 [32]. VEGF receptors are found primarily on endothelial cells, for which VEGF is a potent mitogen [21,26]. VEGF also functions to increase vascular permeability [6]. Angiogenesis is a complex phenomenon [14] and can be induced by a number of cytokines [8,15]. For example, two different pathways of angiogenesis have been demonstrated to depend upon different vascular cell integrins, one associated with bFGF, and the other associated with VEGF

0006-8993 / 02 / $ – see front matter  2002 Elsevier Science B.V. All rights reserved. doi:10.1016/S0006-8993(02)03766-6

196

E.V. Famiglietti et al. / Brain Research 969 (2003) 195–204

[16]. Angiogenesis is a critical factor in embryonic development, wound healing, and tumor growth. Immunocytochemical studies have localized VEGF to human astrocytic neoplasms, as well as to the neurons and astrocytes of normal control brains and brains exhibiting the hallmarks of Alzheimer’s disease [3,7,31]. VEGF is prominent in the brain and retina during neural development and primary angiogenesis, but declines in amount postnatally. In VEGF-deficient, homozygous mouse embryos, blood vessel development is abnormal and a lethal phenotype develops, even in heterozygous mouse embryos [5,11]. In developing rat and cat retinas, primary angiogenesis may involve VEGF synthesized by ¨ astrocytes and Muller cells only [30], but immature neurons of the ganglion cell layer (GCL) can be induced to express VEGF in animal models [29]. In adult retina, VEGF increases in ischemia-associated retinal neovascular diseases [1,23], such as proliferative diabetic retinopathy, and is enhanced in retinal vascular endothelium and vessel walls. Hypoxia is reported to increase VEGF production in retinal endothelial cells and ¨ pericytes, in retinal glial cells, including Muller cells, and in retinal pigmented epithelial cells and ganglion cells, in a reversible manner [2]. A recent study of ischemia in macaque retina [28] reports low but detectable levels of VEGF in the GCL and inner nuclear layer (INL) in normal retinas and demonstrates post-ischemic upregulation of VEGF synthesis only in ganglion cells and neurons of the ¨ INL, without obvious involvement of Muller cells or other glial cells. Previously, we reported the immunocytochemical demonstration of VEGF in retinal neurons, particularly in the GCL of human retina [24], and also prominently in ¨ the inner nuclear layer in amacrine cells and Muller cells [9]. We demonstrate here, as previously reported in brief [9], that VEGF is present in all types of neurons and ¨ Muller glial cells in adult human retina.

2. Methods

2.1. Human tissue Human eyes, generously provided by the Central New York Eye Bank (pairs: n59), were processed 1.5 to 6 h postmortem. The range of ages was 39–86 years and the causes of death were: myocardial infarction (n54), cardiac arrest (n52), respiratory failure (n52), and intracerebral hemorrhage (n51). None had grossly or microscopically evident eye disease. The methods used in this study comply with the Declaration of Helsinki.

2.2. Tissue fixation In seven cases, the anterior portion of the globe and the vitreous humor were removed, and the retinas were dissected free prior to fixation in 0.1 M phosphate-buffered

4% paraformaldehyde for 4 h and then processed for immunocytochemistry or stored overnight in 0.1 M phosphate buffer before processing. In two remaining cases, the globes were incised at the limbus and the whole eyes were fixed in 4% paraformaldehyde, in 0.01 M phosphatebuffered saline (PBS), with 0.2 and 0.1% glutaraldehyde for 7 and 2 h, respectively. In the first of these two cases, the eyes were kept for an extended period in 4% PBSbuffered paraformaldehyde and in the second case for 36 h, after which the latter was stored in PBS prior to retinal dissection. Similar results were obtained in all retinas studied that were sufficiently well-fixed to judge the cellular location of reaction product. Better fixation and satisfactory immunoreactivity was obtained in eyes fixed with a small amount of glutaraldehyde. Most of the figures were derived from the most successful case, and the last described above: the retinas of a 59-year-old female, fixed 6 h after cardiac arrest with 0.1% glutaraldehyde in the initial fixative.

2.3. Immunocytochemistry Whole-mount retinas were processed, using a free-floating method. In addition, 12 mm vertical cryostat sections of central and peripheral retina were processed on Fol’s subbed slides. Whole-mount retinas were incubated with a Protein A-purified polyclonal primary antibody, made in the laboratory of Andrew Baird, against amino acids 1–10 of VEGF (ala-pro-met-ala-glu-gly-gly-gln-lys-tyr), ‘VEGF Ab61’, (1:200 or 1:500) with 0.03% Triton X-100, 0.1% gelatin, and 3% bovine serum albumin (BSA). Whole, free-floating retinas were incubated for 48–72 h at 4 8C on an orbital shaker [18]; sections were incubated for 12–24 h. Antibody specificity for VEGF was confirmed by Western blot analysis with its recombinant antigen. Antibody specificity was further confirmed by the absence of staining in sections of brain tissues including retina that were incubated with the solid-phase VEGF Ab61 affinity column ‘flow-through’ [3]. Co-incubations with a second monoclonal primary antibody (Biogenex) against human antigens: either intermediate filaments, anti-vimentin (VM) (1:500), or the gamma subunit of neuron-specific enolase (NSE) (1:8) were carried out in some experiments. In experiments using DAB, biotinylated goat antirabbit IgG (1:200) was applied for 1 h at room temperature. Then specimens were incubated in streptavidin, conjugated with DAB, and washed in buffer. Negative control sections were processed as above, with omission of the primary antibodies, and these showed low background and no specific staining of neuronal or glial elements (Fig. 1). In double-labeling experiments, using fluorescent secondary antibodies, whole retinal segments or retinal sections were incubated for 5–18 h or 2 h, respectively, in goat antirabbit Cy3 (1:500) and goat anti-mouse FITC (1:75) IgGs (H & L) with sodium azide, 0.1% gelatin, 0.03% Triton X-100, and 1% normal goat serum.

E.V. Famiglietti et al. / Brain Research 969 (2003) 195–204

197

400 daylight film or T-Max 400 film. Dual wavelength laser-scanning confocal microscopy was carried out on a Zeiss 410 confocal microscope with a 633, N.A. 1.4, oil immersion objective lens. Digital images were processed in Adobe Photoshop for contrast and brightness. Green and red pseudocoloring was assigned to anti-VEGF and antiNSE, respectively, in Fig. 2.

3. Results In adult human retina that is free from macroscopic retinal and vascular pathology, VEGF Ab61 labels both neurons and glia. To assist in clarifying the identities of cells labeled, double-fluorescent labeling was carried out with two additional antibodies: to the neuronal marker, ¨ NSE, and the Muller glial cell marker, VM that labels the intermediate filament, vimentin. Fig. 1. VEGF-immunoreactive labeling of a whole-mounted, sector of normal human retina, 2–3 mm from the fovea centralis, processed by the ‘free-floating’ method and visualized with indirect immunocytochemistry, using diaminobenzidine as the chromogen (a). A control sector of retina was processed simultaneously in the absence of primary antibody (b). Arrows indicate the labeling of neuronal cell bodies in the ganglion cell layer, and arrowheads indicate individual cell bodies. Asterisks mark the branching points of superficial blood vessels. Magnification: 334.

2.4. Fluorescence microscopy and image processing In the case of fluorescent labeling, sections were photographed under epifluorescence illumination using Leitz FITC and TRITC filter sets, principally through a 253 microscope objective, exposing Kodak Ektachrome Elite

3.1. VEGF and NSE double labeling In normal human retina, anti-NSE appears to mark neuronal cell bodies of all types, although the intensity of cell body labeling varies considerably from cell-to-cell of the same general class (e.g. ganglion cells). It also marks in a more consistent fashion the processes of these cells, including dendrites in the outer plexiform layer (OPL) and inner plexiform layer (IPL), as well as ganglion cell axons in the optic nerve fiber layer (NFL). In Fig. 2, NSE immunoreactivity (NSE-IR) in the inner half of the retina is illustrated in red. Small cell bodies of amacrine cells at the outer (upper) border of the IPL are labeled (single arrowheads, in Fig. 2a, and asterisks in Fig. 2b), as are large cell bodies of ganglion cells (large arrows, Fig. 2a) in the ganglion cell layer (GCL).

Fig. 2. Double-labeled retina: fluorescence immunocytochemistry for anti-VEGF (green) and anti-NSE (red). NSE immunoreactivity (NSE-IR) labels neuronal elements and smooth muscle cells of vascular walls; VEGF-IR (green) labels neuronal and glial elements, as well as vascular endothelial cells. (a) Retinal location: near periphery. Neuronal cell bodies of the inner nuclear layer (INL) and ganglion cell layer (GCL) are labeled by NSE-IR (red), VEGF-IR (green), or both (yellow). Neuronal processes in the IPL are labeled only by NSE-IR (punctate red labeling). The innermost sublayer of the INL, comprised of amacrine cell bodies, is labeled mostly by VEGF-IR. A few amacrine cell bodies are also labeled by NSE-IR (e.g. single arrowheads). This labeling is mainly cytoplasmic, as can be appreciated in the orange-yellow rim of labeling in the central example (asterisk). In contrast, the VEGF-IR labeling is mainly nuclear (smaller green profiles separated by dark, unlabeled cytoplasm). In the ganglion cell layer, VEGF-IR is also mainly nuclear (single arrows; green or yellow center), whereas NSE-IR of several large-bodied ganglion cells is cytoplasmic (yellow or red periphery). Note the red, ¨ ‘C’-shaped, NSE-IR profile of ganglion cell cytoplasm at the center of the figure. Just above the amacrine cell bodies is a sublayer of Muller glial cell bodies that exhibit nuclear and probably cytoplasmic labeling with VEGF-IR (double arrowheads). A small blood vessel is visible at the right. The innermost (endothelial) layer of the vessel is labeled by VEGF-IR (small arrow), whereas the intermediate, muscular layer is labeled by NSE-IR (small double arrow). (b) and (c) represent double and single (VEGF-IR) labeling of the same field, as in the pair of (d) and (e). Both fields lie in central retina, between the fovea and the optic disc. (b) In more central retina, three bands of NSE-IR neuronal processes can be seen in the IPL (horizontal, white bars). ¨ Three amacrine cell bodies are labeled by NSE-IR (asterisks), the right-hand example (red) exclusively. The row of VEGF-IR Muller cell bodies is clearly seen (double arrowheads). Nuclear VEGF-IR is seen in ganglion cells and displaced amacrine cell bodies in the GCL, but little NSE-IR is detected. NSE-IR ¨ is present in ganglion cell axons, however, collected in bundles in the nerve fiber layer (NFL), surrounded by sheets of (green) Muller cells processes ¨ (dotted circle), which are also interleaved in the bundles (yellow). The VEGF-IR Muller cell end-feet, apposed to the retinal inner limiting membrane, are ¨ seen at the bottom (green). (c) Single labeling with VEGF-IR; upper 2 / 3 of field in (b). Vertically oriented Muller cell processes, rather than laminated ¨ (neuronal) processes, are seen to cross the IPL. The row of smaller Muller cell bodies is easily seen, as in (b) (double arrowheads). (d) Double labeling shows three amacrine cell bodies, yellow at the center with a reddish surround, consistent with nuclear VEGF-IR labeling and cytoplasmic NSE-IR labeling. Other features as noted in (b). (e) Single labeling with VEGF-IR; upper 2 / 3 of field in (d). Nuclear labeling of the three amacrine cell bodies is ¨ ¨ readily seen. Labeled Muller cell bodies are also detected in the row above, as in (d) (double arrowheads). Vertically oriented Muller cell processes are distinct and traverse the IPL (arrows). Magnification: 3900.

198

E.V. Famiglietti et al. / Brain Research 969 (2003) 195–204

E.V. Famiglietti et al. / Brain Research 969 (2003) 195–204

VEGF immunoreactivity (VEGF-IR), illustrated in green (Fig. 2), is present in many circular profiles at the INL / IPL border that by their position and size are the nuclei of amacrine cells (see also Fig. 4). Double labeling (red-green superposition), indicated by yellow, is present in some amacrine cell bodies (single arrowheads) and ganglion cell bodies. A propensity for nuclear VEGF-IR and cytoplasmic NSE-IR can be appreciated in double-labeled cells as yellow centers with red surrounds (right side, Fig. 2a, and asterisks, Fig. 2d), or as green centers with yellow penumbrae (asterisk, and large arrow, left side, Fig. 2a). Punctate NSE-IR labeling, typical of the labeling of synaptic boutons [33] fills the IPL, and in more central retina exhibits a trilaminar banding (white bars in Fig. 2b and d), the central band apparently marking the a / b sublaminar border [10]. Near the optic nerve head (Fig. 2b–d), large bundles of ganglion cell axons (red) are surrounded (green) and intimately invested (yellow) by lamellar processes of ¨ ¨ Muller cells (Fig. 2b and d, bottom). VEGF-IR Muller cell body labeling is conspicuous in these examples as a row of small, angular, polygonal profiles in the expected position, lying one to one-and-a-half rows above the amacrine cell bodies (double arrowheads in Fig. 2). Occasional amacrine ¨ cell bodies that interrupt the row of Muller cell bodies in the second tier show NSE immunoreactivity. These have round, rather than polygonal forms. In optimal sections, delicate VEGF-IR labeling of processes in the IPL can be observed when singly labeled (Fig. 2c and e). The quality of this labeling, unlike the punctate labeling of NSE-IR processes, is generally filamentous. The filamentous processes extend perpendicular to the laminae of the IPL (esp. arrows in Fig. 2e), ¨ following the course of Muller cell processes in the IPL, rather than parallel to the laminae, as do the majority of neuronal processes. The filamentous VEGF-IR labeling of ¨ the IPL is continuous with broader expanses of Muller cell labeling in lamellar processes that enfold the neurons of the ganglion cell layer and invest the axons of the nerve fiber layer. This labeling extends below the axon bundles ¨ (bottom of Fig. 2a and b), lying within the Muller cell end-feet, but is not continuous at the base. Thus VEGF-IR does not label the continuous membranous surface abutting the inner limiting membrane of the retina.

199

optic nerve head. In Fig. 3, slender bundles of VM-IR intermediate filaments are observed to extend perpendicu¨ larly through the retinal layers. The location of Muller cell bodies, noted above and in Fig. 2 to lie just above the amacrine cell sublayer of the INL, cannot be discerned in this preparation. It is evident that the bulk of intermediate ¨ filaments in Muller cells lies in the broad lamellar processes near the inner surface of the retina. In the obliquely-sectioned retina of Fig. 4, VEGF-like immunoreactivity (VEGF-IR) in cells and their processes can be seen in all layers of normal human retina (Fig. 4a). Labeling is prominent in photoreceptor outer segments and the inner segments of rods, while the outer portions (ellipsoid) of the inner segments of cones, rich in mitochondria, stand out in relief as less prominently labeled (arrowheads in Fig. 1a). Close packing of photoreceptor ¨ cell bodies and their investment by Muller cell processes makes it difficult to identify the targets of labeling in the outer nuclear layer (ONL). The majority of neuronal nuclei in the INL are labeled, including most amacrine cells, some of which exhibit the brightest labeling. As shown in Fig. 2a, ganglion cell labeling is present with prominent nuclear labeling (large dot in Fig. 4a). Exclusively nuclear labeling of ganglion cells and displaced amacrine cells is easy to appreciate in the scattered, tangentially sectioned array of these cells (small dots in Fig. 4a). Tangential sectioning of the GCL and NFL with VEGF /

3.2. VEGF and VM double labeling ¨ The great majority of macroglia in the retina are Muller cells that form a highly ordered scaffolding of the retina, extending from the base of the photoreceptor inner segment, where they form the outer limiting membrane, through the retinal layers to the inner surface, where they ¨ form a continuous layer of Muller cell end-feet under the inner limiting membrane. Among neurons and glia, VM-IR ¨ labels only Muller cells in the retina, and presumably the scattered astrocytes that lie at the retina surface near the

Fig. 3. Vimentin (VM)-IR. VM-IR labels the vertical, ascending pro¨ ¨ cesses of Muller cells, from the Muller cell end-feet below the nerve fiber layer (NFL) adjacent to the inner retinal surface, and extending to the outer limiting membrane (OLM), just below the photoreceptor inner ¨ segments. Intermediate filaments are most abundant in the Muller cell end-feet. Dots are centered in lacunae formed by amacrine and ganglion cell bodies labeled by VEGF-IR (not shown) in this doubly labeled ¨ section, marking the boundaries of the IPL. The locations of Muller cell somata, which lie just above the amacrine cells, are not detected in such a preparation. Horizontal elements in the GCL and NFL, as well as the INL, apparently consist of vascular components (v) (Figs. 2a and 6), but also include glial elements (Fig. 5). Magnification: 3520.

200

E.V. Famiglietti et al. / Brain Research 969 (2003) 195–204

Fig. 4. Double-labeled retina. VEGF-IR in (a). VM-IR in (b). (a) An oblique section through the retinal layers, including the cytoplasmically labeled photoreceptors of the outer nuclear layer (ONL), reveals VEGF-IR, prominent in amacrine cells (ACs) at the inner border of the inner nuclear layer (INL), displaced ACs, small ganglion cells (GCs), and large GCs (dots underneath). Note differences in density of label in the large GC (large dot), apparently higher in the nucleus and lower in the nucleolus. Nuclear labeling apparently predominates in amacrine and small ganglion cells. Fibrillar and amorphous labeling in ¨ the GCL and axon layer beneath lies in the region of the Muller cell end-feet, which shows bright VM-IR (right). (b) Note the lacunae formed by the latter, in the which the cell bodies of the GCs (and displaced ACs) lie (centered dots, right). Note that the lacunae around the small GCs, encompassing their cytoplasm, are significantly larger than the nuclear labeling. Some lacunae are empty of VEGF-IR (compare (a) and (b)). Blood vessels (v) show different patterns of labeling with the two markers (Fig. 5). Magnification: 3520.

VM double labeling illustrates circular voids that mark the sites of ganglion cell and displaced amacrine cell bodies and illustrate the investment of neuronal cell bodies in the ¨ ganglion cell layer by the processes of Muller cells (dots in Fig. 4b). A comparison of Fig. 4a and b shows that VM-IR labeling extends to the outer limiting membrane. Moreover, there is partial correspondence between the filamentous VEGF-IR labeling in the IPL and the bundles of ¨ intermediate filaments in Muller cells (left side, Fig. 4a and b). VEGF-IR labeling in the NFL, although much less ¨ prominent that that of intermediate filaments in Muller

cells, shows partial overlap with VM-IR labeling. The filamentous nature of the VM-IR in the GCL and NFL is easily appreciated (Fig. 4b), but there is also some apparently filamentous VEGF-IR focally seeming to follow similar paths (Fig. 4a). It is not obvious from these patterns whether the VEGF-IR here labels ganglion cell axons or ¨ Muller cell end-feet (cf. Fig. 2). To examine this correspondence more closely, retinas double labeled for VEGF and VM were examined by confocal microscopy, as illustrated in Fig. 5. Laser confocal scanning images of the NFL, were taken

E.V. Famiglietti et al. / Brain Research 969 (2003) 195–204

201

Fig. 5. Double-labeled retina. VEGF-IR in (a). VM-IR in (b). Laser confocal scanning microscopy of whole-mounted retina, 3.5 mm deep to the retinal surface beneath the inner limiting membrane, shows similar but not identical patterns of horizontally organized fibrillar VM-IR and fibril-associated ¨ VEGF-IR. Comparison with Fig. 4 shows that the fibrils are distributed throughout the NFL in a disorganized network indicative of localization in Muller cell processes that interdigitate among neuronal elements. The smaller voids correspond to the positions of unlabeled ganglion and displaced amacrine cell bodies (cf. Fig. 4). No evidence for labeling of ganglion cell axons is seen. Tangentially oriented vascular elements (running vertically) are detectable by VEGF-IR. Imaged with a 633, N.A. 1.4, oil immersion objective lens. Magnification: 3510.

parallel to and a few microns deep to the retinal surface of a whole-mounted retina successfully labeled with antibodies both to VEGF and VM (Fig. 5a and b, respectively). It is clear that while the two patterns of labeling are not identical, VEGF-IR closely follows the distribution of ¨ intermediate filaments in Muller cell end-feet. The VEGFIR is somewhat less extensive than VM-IR and does not label other elements in a discrete fashion. Thus there is no evidence for ganglion cell axonal labeling in addition to ¨ Muller cell labeling.

3.3. Vascular labeling by VEGF-IR

Fig. 6. Double-labeled retina. VEGF-IR in (a). VM-IR in (b). Two layers of the blood vessel (v) wall are VEGF-IR (arrowheads). VM-IR labels the intermediate filaments of endothelial cells forming the innermost layer. (The inner portion of the outer layer is labeled by anti-NSE in Fig. 2a.) ¨ Vessels are bounded by VM-IR Muller cell processes. Dots are placed under a labeled ganglion cell body in (a) and in the center of the VM-IR void formed by the same cell body in (b). Magnification: 3500.

As expected, vascular VEGF-IR is consistently present in all preparations. In the largest vessels, a double layering of labeling can be visualized (arrowheads in Fig. 6a). The inner band of labeling is coincident with delicate VM-IR labeling of intermediate filaments in endothelial cells [27], demarcating the vascular lumen (arrowheads in Fig. 6b). In addition, VEGF-IR labels an outer layer of the arteriolar vessel that includes the vascular media [13]. Note that ¨ vessels are surrounded by Muller cell processes and endfeet in the NFL (Fig. 6b). In a comparison of vascular labeling by NSE-IR and VEGF-IR, multilaminar VEGF-IR labeling is also seen in

202

E.V. Famiglietti et al. / Brain Research 969 (2003) 195–204

Fig. 2a. The innermost, more prominent endothelial layer is labeled by VEGF-IR (green, small arrow in Fig. 2a). The layer labeled by NSE-IR is external to the endothelium (red, double arrow in Fig. 2a), and apparently consistent with the muscular media [20].

4. Discussion We have demonstrated that neurons, in particular amacrine and ganglion cells, are immunoreactive for a polyclonal antibody against the amino-terminal portion of the human VEGF molecule in adult human retinas that are free of clinical or microscopic evidence of disease. In ganglion cells, labeling is typically most intense in the nucleus, but is also present in the cytoplasm. The labeling of amacrine cells is similar to that of ganglion cells, with cytoplasmic labeling greater in some cells, but nuclear labeling is predominant. We were unable to demonstrate unequivocal VEGF-IR in the dendrites of amacrine and ganglion cells and in ganglion cell axons, as demonstrated in double labeling with anti-NSE, which reveals horizontal bands of dendrites in the IPL and bundles of axons in the nerve fiber layer, absent in VEGF-IR. The labeling of horizontal cells and bipolar cells of the inner nuclear layer is diffuse and in the latter case less intense. Delicate, vertically oriented VEGF-IR processes in the IPL had the appearance of the ¨ axial component of Muller cell glial processes, rather than the more robust and partially stratified axonal systems of bipolar cells. There is evidently some VEGF-IR in photoreceptors, most clearly seen in rod inner and outer segments (Fig. 4a). The observed outlining of rod nuclei could be the ¨ labeling of Muller cell processes only, but the labeling of rod inner segments suggests that this may include cytoplasmic rod labeling. There is labeling of the outer plexiform layer, but we did not carry out high resolution imaging of the OPL, because fixation and freezing artefacts would raise questions about the integrity and identity of the fine neuronal processes intermingled in the OPL that belong to several different types of cell in that location, even if electron microscopy were used. Gerhardinger and coworkers [17], also using an antibody to the amino-terminal portion of VEGF, common to all isoforms present in retina, compared their results with those we set forth in a previous brief report [25], which were comparable in regard to neuronal labeling. In common with our materials and methods, their postmortem intervals were also relatively short. Gerhardinger and coworkers [17] also carried out in situ hybridization demonstrating synthesis of VEGF with variable intensity in the INL and GCL of both normal and diabetic eyes. While these authors did not specifically acknowledge the strong immunocytochemical labeling of amacrine cell bodies we found in both the INL and the GCL, they provided convincing evidence of cytoplasmic labeling in large

neurons consistent with ganglion cell bodies in the GCL. They did not find consistent differences between normal and diabetic eyes. Although these authors identified labeling of the inner and outer limiting membranes, they made ¨ no mention of the Muller cell labeling that we found in our study, and that was implied by the pattern they illustrated in their study. The vascular labeling by anti-VEGF that we found in the location of endothelial cells could be the result of binding at VEGF receptors with internalization into endothelial cell cytoplasm, or synthesis within endothelial cells. ¨ We have also demonstrated the labeling of Muller glial cells by anti-VEGF. Examination of the axon layer and inner limiting membrane by laser confocal microscopy in a double-labeled comparison of VEGF-IR and vimentin-IR, ¨ ¨ marking Muller cells makes it clear that Muller cell labeling is present in normal human retina. The labeling of ¨ Muller cells is somewhat variable from one location to another in our preparations and does not always distinctly show the vertical processes passing through the inner plexiform layer, illustrated in Figs. 2–4, the visualization of which may also depend upon achieving an optimal angle of tissue sectioning of the retinal layers, as well as optimal tissue preservation. ¨ Muller cell labeling here is much less prominent than that observed in diabetic eyes by Amin and coworkers [4], who showed a pattern of VEGF-IR labeling quite similar to the pattern of vimentin labeling throughout the retinal layers. In their examination of subjects with nonprolifera¨ tive diabetic retinopathy, Muller cell labeling with antiVEGF was very similar to the results of our study in the close association of the VEGF-IR pattern with the vimentin-IR pattern of intermediate filament staining. Our results differed, however in also showing label in the cell ¨ bodies of Muller cells, whereas in their diabetic retinas cell body labeling is not shown despite intense fibrillar labeling throughout the retinal layers (see Fig. 3A in Ref. [4]). Disease had evidently caused pathological changes in these ¨ Muller cells, however, as they were also strongly immunoreactive for GFAP. Another immunocytochemical study of VEGF in normal and diabetic eyes found weak or no labeling of any kind in normal eyes [22]. In both of those studies, however, postmortem intervals were significantly longer, ranging from 29 to 36 h in the study of Amin and coworkers [4], whereas our specimens were obtained no longer than 8 h postmortem. Another possible explanation for the absense of neuronal labeling in the previously cited studies is the use of a different antibody to VEGF. That used in our study was made against the amino-terminal region of VEGF-IR, common to all isoforms of the molecule. Evidence from molecular cloning of VEGF in macaque retina [28], indicates that the two diffusible forms of VEGF (VEGF 121 and VEGF 165 ) are significantly upregulated in experimental retina ischemia, but VEGF 121 predominates in nonischemic retina. The antibody used by Amin and coworkers

E.V. Famiglietti et al. / Brain Research 969 (2003) 195–204

[4] is made against the whole VEGF 165 molecule, and cross-reacts only slightly with VEGF 121 (R&D Systems, Inc., personal communication, A. Fleming). Assuming similar expression of the nearly identical VEGF isoforms in macaque and human retinas, the antibody used by Amin and coworkers [4] might not have detected VEGF in human subjects without retinal disease, as they reported, even though VEGF 121 was present and apparently detectable by the antibody we used. It is also possible that VEGF 121 and VEGF 165 are differently distributed (the latter has moderate heparin binding affinity), when they are upregulated in experimental or diabetic ischemia, and that secreted diffusible VEGF 121 has a nuclear and cytoplasmic localization with in addition a limited tendency to associate with intermediate filaments, whereas VEGF 165 tends to colocalize with intermediate filaments, predominantly in glial cells. It was the conclusion of Shima and coworkers [28] in their experimental study of macaque retina, using a high resolution method of in situ localization of mRNA, that ganglion cells significantly upregulated their synthesis of VEGF in response to hypoxia (which might also be expected to occur in the immediate postmortem interval). They acknowledged, however, that their method was too insensitive to detect low levels of VEGF expression in ¨ normal retina, nor was their method likely to detect Muller cell body labeling under any circumstances, given the high level of labeling of other, adjacent and larger cell bodies in the INL. As we have shown with an antibody that presumably detects all isoforms of VEGF in the retina, VEGF is present in the retinas of human adults that are free of clinical eye disease and free of retinal pathology by gross and microscopic inspection. If the synthesis of the freely diffusible form(s) of VEGF occurs in retinal neurons, constitutively and at low levels in normal retina and at increased levels in hypoxic conditions, as suggested by the work of Shima and coworkers [28], then neurons of the inner retina are potentially endowed with the ability to regulate their own blood supply. They may also be able to modulate their own output of cytokines by feedback regulation of transcription. The role that glial cells play in the regulation of vascular supply may be distinct from that of neurons and of equal or greater importance in disorders such as diabetic retinopathy [4]. It is evident that VEGF in glial cells plays important roles in other circumstances, both normal and pathological, as in the development of retinal vasculature [30] and in the proliferation of vessels in glial neoplasms of the brain [3].

Acknowledgements The authors are grateful to Dr Andrew Baird and A.M. Gonzales for providing the antibody to VEGF. E.D. McGookin was supported by a Juvenile Diabetes Associa-

203

tion summer fellowship. This work was supported by National Institute of Health grants AG10682 (E.G.S.) and EY01602 (B.W.S.).

References [1] L.P. Aiello, R.L. Avery, P.G. Arrigg, B.A. Keyt, H.D. Jampel, S.T. Shah, L.R. Pasquale, H. Thieme, M.A. Iwamoto, J.E. Park et al., Vascular endothelial growth factor in ocular fluid of patients with diabetic retinopathy and other retinal disorders, N. Engl. J. Med. 331 (1994) 1480–1487. [2] L.P. Aiello, J.M. Northrup, B.A. Keyt, H. Takagi, M.A. Iwamoto, Hypoxic regulation of vascular endothelial growth factor in retinal cells, Arch. Ophthalmol. 113 (1995) 1538–1544. [3] J.A. Alvarez, A. Baird, A. Tatum, J. Daucher, R. Chorsky, A.M. Gonzalez, E.G. Stopa, Localization of basic fibroblast growth factor and vascular endothelial growth factor in human glial neoplasms, Mod. Pathol. 5 (1992) 303–307. [4] R.H. Amin, R.N. Frank, A. Kennedy, D. Eliott, J.E. Puklin, G.W. Abrams, Vascular endothelial growth factor is present in glial cells of the retina and optic nerve of human subjects with nonproliferative diabetic retinopathy, Invest. Ophthalmol. Vis. Sci. 38 (1997) 36–47. [5] P. Carmeliet, V. Ferreira, G. Breier, S. Pollefeyt, L. Kieckens, M. Gertsenstein, M. Fahrig, A. Vandenhoeck, K. Harpal, C. Eberhardt, C. Declercq, J. Pawling, L. Moons, D. Collen, W. Risau, A. Nagy, Abnormal blood vessel development and lethality in embryos lacking a single VEGF allele, Nature 380 (1996) 435–439. [6] D.T. Connolly, Vascular permeability factor: a unique regulator of blood vessel function, J. Cell. Biochem. 47 (1991) 219–223. [7] R. Corona, R. Chorsky, A. Baird, A. Gonzalez, L. Patel, E. Stopa, Vascular endothelial growth factor: a heparin-binding growth factor with a different distribution than bFGF in normal brain and Alzheimer’s disease, Soc. Neurosci. Abstr. 83 (1991) 196. [8] P.A. D’Amore, R.W. Thompson, Mechanisms of angiogenesis, Annu. Rev. Physiol. 49 (1987) 453–464. [9] E. Famiglietti, P. Song, E. McGookin, B. Streeten, V. Kuo-Leblanc, A. Baird, A. Gonzalez, E. Stopa, Immunocytochemical localization of vascular endothelial growth factor in neurons and glia of human retina, Soc. Neurosci. Abstr. 22 (1996) 2017. [10] E.V. Famiglietti Jr., H. Kolb, Structural basis for ON- and OFFcenter responses in retinal ganglion cells, Science 194 (1976) 193–195. [11] N. Ferrara, K. Carver-Moore, H. Chen, M. Dowd, L. Lu, K.S. O’Shea, L. Powell-Braxton, K.J. Hillan, M.W. Moore, Heterozygous embryonic lethality induced by targeted inactivation of the VEGF gene, Nature 380 (1996) 439–442. [12] N. Ferrara, W.J. Henzel, Pituitary follicular cells secrete a novel heparin-binding growth factor specific for vascular endothelial cells, Biochem. Biophys. Res. Commun. 161 (1989) 851–858. [13] N. Ferrara, J. Winer, T. Burton, Aortic smooth muscle cells express and secrete vascular endothelial growth factor, Growth Factors 5 (1991) 141–148. [14] J. Folkman, M. Klagsbrun, Angiogenic factors, Science 235 (1987) 442–447. [15] J. Folkman, Y. Shing, Angiogenesis, J. Biol. Chem. 267 (1992) 10931–10934. [16] M. Friedlander, P.C. Brooks, R.W. Shaffer, C.M. Kincaid, J.A. Varner, D.A. Cheresh, Definition of two angiogenic pathways by distinct alpha v integrins, Science 270 (1995) 1500–1502. [17] C. Gerhardinger, L.F. Brown, S. Roy, M. Mizutani, C.L. Zucker, M. Lorenzi, Expression of vascular endothelial growth factor in the human retina and in nonproliferative diabetic retinopathy, Am. J. Pathol. 152 (1998) 1453–1462. [18] A.M. Gonzalez, M. Buscaglia, M. Ong, A. Baird, Distribution of

204

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

E.V. Famiglietti et al. / Brain Research 969 (2003) 195–204 basic fibroblast growth factor in the 18-day rat fetus: localization in the basement membranes of diverse tissues, J. Cell Biol. 110 (1990) 753–765. D. Gospodarowicz, J.A. Abraham, J. Schilling, Isolation and characterization of a vascular endothelial cell mitogen produced by pituitary-derived folliculo stellate cells, Proc. Natl. Acad. Sci. USA 86 (1989) 7311–7315. H. Haimoto, Y. Takahashi, T. Koshikawa, H. Nagura, K. Kato, Immunohistochemical localization of gamma-enolase in normal human tissues other than nervous and neuroendocrine tissues, Lab. Invest. 52 (1985) 257–263. D.W. Leung, G. Cachianes, W.J. Kuang, D.V. Goeddel, N. Ferrara, Vascular endothelial growth factor is a secreted angiogenic mitogen, Science 246 (1989) 1306–1309. G.A. Lutty, D.S. McLeod, C. Merges, A. Diggs, J. Plouet, Localization of vascular endothelial growth factor in human retina and choroid, Arch. Ophthalmol. 114 (1996) 971–977. F. Malecaze, S. Clamens, V. Simorre-Pinatel, A. Mathis, P. Chollet, C. Favard, F. Bayard, J. Plouet, Detection of vascular endothelial growth factor messenger RNA and vascular endothelial growth factor-like activity in proliferative diabetic retinopathy, Arch. Ophthalmol. 112 (1994) 1476–1482. E. McGookin, E. Stopa, R. Chorsky, G. Collins, A. Baird, A. Gonzalez, Vascular endothelial growth factor (VEGF) in normal and diabetic human retina, Invest. Ophthalmol. Vis. Sci. 33 (1992) 821. E. McGookin, E. Stopa, V. Kuo-Leblanc, A. Baird, A. Gonzalez, A. Hanneken, B. Streeten, Vascular endothelial growth factor (VEGF) has a different distribution than basic fibroblast growth factor (bFGF) in the adult human retina, Invest. Ophthalmol. Vis. Sci. 33 (1992) 821. J. Plouet, J. Schilling, D. Gospodarowicz, Isolation and characteriza-

[27]

[28]

[29]

[30]

[31]

[32]

[33]

tion of a newly identified endothelial cell mitogen produced by AtT-20 cells, EMBO J. 8 (1989) 3801–3806. F. Ramaekers, G. Vooijs, A. Huijsmans et al., Immunohistochemistry as an aid in diagnostic cytopathology, in: R. DeLellis (Ed.), Advances in Immunohistochemistry, Raven Press, New York, 1988, pp. 133–163. D.T. Shima, A. Gougos, J.W. Miller, M. Tolentino, G. Robinson, A.P. Adamis, P.A. D’Amore, Cloning and mRNA expression of vascular endothelial growth factor in ischemic retinas of Macaca fascicularis, Invest. Ophthalmol. Vis. Sci. 37 (1996) 1334–1340. J. Stone, T. Chan-Ling, J. Pe’er, A. Itin, H. Gnessin, E. Keshet, Roles of vascular endothelial growth factor and astrocyte degeneration in the genesis of retinopathy of prematurity, Invest. Ophthalmol. Vis. Sci. 37 (1996) 290–299. J. Stone, A. Itin, T. Alon, J. Pe’er, H. Gnessin, T. Chan-Ling, E. Keshet, Development of retinal vasculature is mediated by hypoxiainduced vascular endothelial growth factor (VEGF) expression by neuroglia, J. Neurosci. 15 (1995) 4738–4747. E. Stopa, J. Alvarez, R. Corona, R. Chorsky, A. Baird, A. Gonzalez, Vascular endothelial growth factor: yet another growth factor produced by astrocytic neoplasms, J. Neuropathol. Exp. Neurol. 50 (1991) 291. E. Tischer, R. Mitchell, T. Hartman, M. Silva, D. Gospodarowicz, J.C. Fiddes, J.A. Abraham, The human gene for vascular endothelial growth factor. Multiple protein forms are encoded through alternative exon splicing, J. Biol. Chem. 266 (1991) 11947–11954. J.E. Vaughn, E.V. Famiglietti Jr., R.P. Barber, K. Saito, E. Roberts, C.E. Ribak, GABAergic amacrine cells in rat retina: immunocytochemical identification and synaptic connectivity, J. Comp. Neurol. 197 (1981) 113–127.