Microvascular Remodeling after Occlusion-Recanalization of a Branch Retinal Vein in Rats

Abstract

purpose. To describe the time course of microvascular changes after transient branch retinal vein occlusion (BRVO) in rats.

methods. BRVO was induced in pigmented rats by focal laser photocoagulation. The subsequent changes in the retinal angiogram were followed up, both in vivo by confocal scanning laser ophthalmoscopy and ex vivo by confocal microscopy.

results. At day 1, capillary closure affected the three microvessel layer differentially, the intermediary layer being the most affected. Collateral veins, which were initiated by the dilation of deep-layer venules, pursued their course below adjacent arteries. These microvascular changes peaked between days 1 and 3. After recanalization at day 3, microvascular changes regressed gradually but incompletely, and at day 30 capillary closure and venule dilation persisted.

conclusions. Transient occlusion of a retinal vein in rats leads to short- and long-term microvascular remodeling upstream of the occlusion site. This study describes a model for the tridimensional arrangement of retinal microvessel that accounts for the topography of the early capillary closure and collateral vessel formation that occur after BRVO. In the long term, these changes regressed incompletely, with recanalization of the occluded vein, suggesting that after a short period of occlusion, microvascular changes may become at least partially independent of flow. Despite the intrinsically limited applicability of this model to human vein occlusion, the results suggest that even if therapeutic decompression of an occluded vein is performed early, it may not reverse capillary dropout completely.

Retinal vein occlusion leads to a complex evolutive pattern of microvascular remodeling, with a time course and pathophysiology that are still unclear. In animals, the earliest microvascular changes after branch retinal vein occlusion (BRVO) include venous dilation, decreased and/or reversed venous flow, ^{1 2} arterial constriction, ^{3 4} and increased venous pressure. ⁵ These changes are followed by capillary dropout, collateral vessel formation, and eventually neovascular proliferation. ^{6 7} However, because of the technical limitations of fluorescein angiography, much remains to be known about the time course and quantification of these microvascular changes. Detailed in vivo analysis of microvascular perfusion is out of reach of conventional fluorescein angiography, because the latter does not allow accurate visualization, either of the entire retinal microvasculature, especially the deep layer, ⁸ or of its spatial arrangement. Therefore, techniques that allow optical sectioning and tridimensional reconstruction are preferable. Confocal scanning laser ophthalmoscopy (cSLO) and confocal microscopy have been applied only recently to the study of the retinal microvasculature. ^{9 10} In the present study, we used cSLO and confocal microscopy to establish in detail the time course of the microcirculatory changes that occur after transient occlusion of a single retinal vein in rats. 

Animals and Methods

All manipulations were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Adult male Brown Norway rats were provided by Janvier (Saint-Ile le Genest, France). Anesthesia was induced by intraperitoneal injection of 100 mg/kg ketamine and 25 mg/kg xylazine (both from Sigma-Aldrich, Lyon, France). For fluorescein angiography and laser photocoagulation, 0.1 mL of 10% sodium fluorescein (Bausch & Lomb, France) was injected intraperitoneally. 

Retinal vein occlusion was performed by laser photocoagulation (wavelength 534 nm; Crystal focus; Alcon Corp., Forth Worth, TX) in anesthetized rats. After pupil dilation with tropicamide (CibaVision, Duluth, GA), the rat was placed in front of a slit lamp. The right eye was chosen as the experimental eye. The fundus was visualized through a SuperPupil lens (Volk) using a slit lamp. The laser spot (duration: 0.5 second; power: 100–150 mW; spot size: 50 μm) was placed on a superior retinal vein, approximately 4 disc diameters from the disc. Care was taken to obtain accurate focusing of the aiming beam on the vessel by adjusting the distance between the lens and the eye. Two to five impacts were sufficient to obtain complete occlusion, which appeared to be due to a combination of intravascular thrombosis and external compression of the vein by perivenous retinal edema. Immediately after laser treatment, the fundus was examined by cSLO (Heidelberg Engineering, Heidelberg, Germany). The occlusion was considered complete when there was evidence of slow reversed flow in major veins and dye leakage into the retina. When five laser impacts did not result in occlusion or if intravitreal hemorrhage occurred, the animal was excluded. The sham group (n = 12) consisted of eyes treated with laser impacts outside vessels, with the same parameters and number of impacts as used in experimental animals. During the days after this treatment, the fundus was examined by slit lamp and cSLO. Manual handling of conscious animals allowed easy examination of the fundus. Animals in which a bullous serous retinal detachment developed were excluded. This was often the case in our preliminary experiments, in which the occlusion was performed closer to the disc. The presence of collateral circulation at the border of the occluded territory was appreciated. In areas of interest, serial optical sections were acquired by performing serial z-scans throughout the retinal thickness. 

For confocal microscopy, flatmounted fluorescein dextran-filled retinas were prepared as follows: Plasma was labeled in anesthetized rats by intracavernous injection of 0.2 mL of FITC-dextran (molecular mass: 2000 kDa, 30 mg in 1 mL of phosphate-buffered saline), according to a modified version of a previously reported method. ^{10 11} To ensure the presence of the dye in the capillaries, the fluorescence of iris capillaries was checked under an epifluorescence microscope in the seconds after the injection. Rats were then humanely killed by an overdose of pentobarbital and enucleated. The short period between dye injection and death ensured minimal dye leakage into the retinal tissue. The globes were then placed for 10 minutes in 4% paraformaldehyde. Next, the sclera and retinal pigment epithelium were stripped off with surgical instruments under a binocular microscope, and the retina was fixed for 24 hours. It was then flatmounted with gelatin-glycerol. Six eyes were examined at each time point (1, 3, 8, and 30 days after vein occlusion). Normal, untreated contralateral eyes (n = 8) were also examined as the control. Confocal microscopy was then performed with a microscope (model 500 MRC; BioRad, Hercules, CA) equipped with argon-krypton laser. A retinal area approximately 500 μm upstream of the laser site under the occluded vein was chosen for examination. In each field examined, successive scans were performed through the retina with a ×10 or ×40 lens. The step in the z-axis between scans varied from 0.5 to 5 μm. In the superficial layer, arteries and veins were identified by their branching patterns. The superficial arterioles exhibited numerous dichotomous branchings in the retinal plane before suddenly changing direction to perpendicular, whereas the superficial veins received venules arising from the deep layer directly into their main trunk or through a short oblique path. The vertical capillaries joining the three microvessel layers were identified on high-magnification images with the ×40 lens. For image analysis, images were converted to the TIFF format, using image conversion software (Graphic Converter; Lemke Software, Peine, Germany). The definition of each image was 768 × 512 pixels. The vascular density of each layer of microvessels was calculated on ×10 images. ^{10 12} Vessels were manually outlined with image-analysis software (Photoshop 4.0; Adobe Corp., Mountain View, CA). NIH Image 1.62 Software (available by ftp at zippy.nimh.nih.gov/ or at http://rsb.info.nih.gov/nih-image; developed by Wayne Rasband, National Institutes of Health, Bethesda, MD) allowed the drawings to be converted into their central axes one pixel wide (skeletonization) and also enabled the number of residual pixels to be counted. This number was proportional to the length of the blood vessels. The corresponding surface was measured in square pixels. Capillary and venule diameters were measured on high-magnification (×40) images, using NIH Image software. In addition, laser-treated retinal samples (taken on days 8 and 30 after laser application, n = 3 each) were fixed in paraformaldehyde and embedded in paraffin for light microscope examination. 

Results

Organization of Microvessels in the Rat Retina

By high-magnification confocal microscopy, the course of individual capillaries could be followed within the retinal thickness in successive z-scans, and therefore the arteriolar, capillary, and venular pathways could be delineated. The trilaminar organization of the rat retinal microvessels was clearly visible, as shown in Figures 1 and 2 . Analysis of the capillary connections between the layers allowed the modeling of the tridimensional arrangement of the microvessels. In the following descriptions, “superficial” means close to the vitreous side; “deep,” close to the photoreceptor side; “horizontal,” parallel to the retinal surface; and “vertical,” parallel to the light path. The horizontal arterioles in the superficial retina (Fig. 1A) abruptly made a 90° turn to join the intermediary layer through short vertical capillaries (Fig. 1B) . The intermediary layer was made up of horizontal capillary segments that did not form a complete anastomotic array throughout the retina (Fig. 1C) . The flow then joined the deep layer again through short vertical capillaries, with relatively few capillaries driving the flow from the intermediary layer to the superficial layer. In the deep layer, a dense capillary plexus coexisted with postcapillary venules (Fig. 1D) . These venules abruptly changed direction to vertical, to join a major vein in the inner retina (Fig. 2) . The confluences of these transverse venules under major veins were 150 to 200 μm away from each other. Overall, most of the anatomic connections drove the arteriolar flow from the superficial arteries to the deep microvessel layers, and vice versa for the venular flow. This pattern of parallel but opposed arteriolar and venular flows coexisted with capillary pathways that shunted the deep layer—that is, with superficial-layer capillaries that branched directly into major veins, such as those shown in Figure 2A , and with capillary segments connecting the intermediary layer to the major veins. 

The confocal scans of fluorescein angiograms obtained by cSLO allowed us to distinguish between the superficial and deep microvessel layers (Figs. 1E 1F) . The venules in the deep layer were clearly visible, but neither the intermediary microvessel layer nor the vertical connecting capillaries could be clearly distinguished. 

Microvascular Changes after BRVO

Immediately after BRVO, dilation of the upstream venous segment and constriction of the downstream segment were noted, together with dye leakage from deep capillaries (Fig. 3) and reversed flow upstream of the occlusion site toward the periphery and the neighboring venous territories (data not shown). Deep-layer venules exhibited dilation and increased tortuosity in both the affected and the adjacent normal retina. This led to the formation of collateral vessels at the border of the occluded territory. The course of these vessels ran under the superficial arteries (Fig. 4) . Within hours of occlusion, retinal hemorrhages occurred, and in some cases peripheral serous retinal detachment. By day 3, all rats had recanalized their occluded veins. At the occlusion site, the residual stenosis at day 30 was in all cases less than 30%—that is, the final diameter at the laser site was at least 70% of the upstream venous diameter (Fig. 5) . Concomitantly with recanalization, fluorescein leakage resolved as deep-layer venule tortuosity gradually decreased. However, at day 30, venule dilation was still detectable in both the occluded and adjacent territories, as is illustrated in Figure 6 , which shows a major vein adjacent to the occluded vein. In this image, the venules on the opposite side are normal. 

Capillary density measurements by confocal microscopy revealed a marked decrease in the density of the intermediary and deep-layer capillaries by day 1 after occlusion, followed by progressive but partial correction by day 30 (Fig. 7) . In the superficial layer, capillary density had decreased significantly compared with the control at days 3 and 8, but much less than in the other layers. Examples of confocal microscopy angiograms at days 1 and 30 are shown in Figure 8 . At day 30, the decreases in capillary density compared with the control were respectively 10%, 57%, and 15% in the superficial, intermediary and deep layers (P = ns, < 0.0001, and < 0.0001, respectively, Mann-Whitney test). Venule diameter increased by 32% (P < 0.001), and capillary diameter, by 11% (P < 0.01). By light microscope examination, retinal morphology appeared normal upstream of the laser treatment (Fig. 9) . Sham-treated eyes did not exhibit any of these changes. 

Discussion

In the present study, we investigated the pattern of microvascular changes occurring after transient BRVO in rats. We began by determining the normal microvascular architecture and the relationship between the three microvessel layers, which have not yet been determined in the rat. The interconnection between these layers and therefore the blood distribution were consistent with the presence of a predominantly serial organization of flow—that is, an arteriolar flow running from the superficial to the deep layer, and a venular flow in the opposite direction, from the deep layer to the superficial veins. In particular, there were few links between the intermediary layer and the vertical venules that drain the venous blood from the deep layer up to the major veins, which indicates that most of the major veins in the inner retina are connected to the deep layer. From these results, we propose a scheme of blood flow routing within the normal retina (Fig. 10 , top). This description of the tridimensional arrangement of retinal capillaries completes the findings of other studies in which corrosion cast techniques were used. ^{13 14} However, it should be remembered that the respective proportions of flow routed through the different layers of the retina cannot be deduced solely from these anatomic considerations, because variations in vessel impedance may compensate for variations in capillary length. In addition, there may be active regulation of flow within the retina by the modulation of vessel diameter, as suggested by others. ¹⁵ Thus, the model we have proposed for the retinal vessel organization in the mouse ¹⁰ also appears to be applicable to the rat. 

Despite these limitations, this model of retinal angioarchitecture helps to clarify some of the earliest consequences of BRVO. Early capillary closure indeed predominantly affected the intermediary layer and was also present to a lesser degree in the deep layer. We suggest that when, after BRVO, the flow from the superficial layer arrives in the intermediary layer, it is preferentially directed toward the side of the adjacent nonoccluded venous territory. Consequently, nonperfusion occurs in capillary segments oriented toward the occluded territory. Early capillary closure in the intermediary layer may thus be a hemodynamic process, because the intermediary layer acts as a crossroads. Conversely, the presence of a retrograde flow in major veins, which implies a retrograde flow in the vertical transverse venules and subsequently in the deep layer venules, may keep the deep layer capillaries open and thus prevent capillary dropout in that layer. This hypothesis is illustrated in Figure 10 , bottom. 

In a previous study in mice, we reported that in the hours after branch vein occlusion there was no detectable capillary closure, despite a similar pattern of interlayer connections. ¹⁰ The origin of this discordance is not yet clear. Technical differences may be involved: Here, for instance, we used fluorescein to perform vein occlusion, which results in more efficient occlusion. Also, in mice, due to the extremely small volume of the vitreous, the increase in retinal volume due to laser impacts increases intraocular pressure, which helps to arrest venous flow independently of the occlusion itself. Therefore, venous flow may resume as soon as the intraocular pressure diminishes. Additional studies are underway to compare the consequences of vein occlusion in rats and mice. 

In the present investigation, the deep microvessel layer was electively affected by the venule dilation and dye leakage that immediately follow BRVO, because of the close relationship between the deep-layer venules and superficial veins, due to the presence of vertical transverse venules. We showed in this study that the collateral circulation arises from the dilation of deep layer venules on both sides of the adjacent major arteries and pursues its course under these arteries. The shortest capillary path between two deep-layer venules facing each other on the border of the occluded territory determines the actual pathway of a given collateral vessel. 

The occluded vein began to reopen at day 3, and subsequently, progressive enlargement of the vein at the laser site was observed. At day 30, however, a variable degree of residual stenosis was still present in most rats. Concomitantly, the microvascular changes upstream of the laser site progressively improved but did not return to normal values. We postulate that the residual capillary closure was not due to retinal atrophy consecutive to the initial serous retinal detachment, because this detachment was of short duration, and there was no histologic evidence of retinal atrophy in the affected retina. Moreover, other histologic studies ^{4 6} have shown that post-vein-occlusion retinal atrophy primarily affects the ganglion cells—the level of the superficial microvessel layer—which changed the least in our rats. Therefore, we believe that the observed microvascular changes are a direct consequence of the vein occlusion. The delay between recanalization and improvement of the microcirculation indicates that the microvascular changes are at least partly independent of flow. Microvascular changes secondary to vein occlusion are known to involve complex evolutionary phenomenons that are not well understood. The chronic venous dilation that we found may be related to structural changes in the venule wall, such as pericyte and endothelial cell proliferation, and an increase in collagen content. ^{7 16 17} It is known that during the course of chronic vein occlusion capillary closure may progress over months despite the absence of noticeable change in the venous circulation. ^{7 16} Moreover, extensive capillary nonperfusion may be present despite recanalization of the occluded vein. ⁶ Capillary closure may result from several factors, including arterial perfusion changes, ^{3 4} glial proliferation, ischemia-reperfusion lesions, intraluminal thrombosis, internal compression due to hemorrhage into the basal membrane, ¹⁸ or vascular endothelial growth factor-induced edema of endothelial cells. ¹⁹ Our data provide a likely explanation for the presence of early capillary closure, but do not allow further progress in the understanding of chronic nonperfusion. 

Little is yet known about the functional and metabolic impairment of vein occlusion. The topography of capillary dropout suggests that the inner plexiform layer may have been the area most affected in both the acute and the chronic phase of BRVO. From measurements of the oxygen concentration in the rat retina, Cringle et al. ²⁰ have suggested that in the normal state, the inner plexiform layer has a high oxygen consumption rate, which contrasts with its capillary paucity and may therefore be affected early by a decrease in retinal perfusion. Therefore, it is likely that a permanent decrease in capillary density affects the neural metabolism in that layer. However, in addition to capillary dropout, several factors may contribute to lower retinal oxygen tension in the acute phase of vein occlusion. After BRVO, there is indeed a slow reverse flow in the deep layer, indicating poor perfusion with desaturated blood, which itself probably causes retinal hypoxia. It is also likely that a serous retinal detachment, even if shallow, impairs oxygen delivery to the outer retina. Therefore, retinal ischemia is probably more extensive through the retinal thickness than the topography of capillary closure might suggest. To our knowledge, retinal oxygen levels during vein occlusion has not been measured in rats. In pig and monkey models of BRVO, ^{6 21} Pournaras et al. ²² found reduced levels of preretinal oxygen over nonperfused areas, and transretinal Po ₂ profiles exhibited minimum oxygen levels within the retina. However, the precise level of the minimum Po ₂ was not specified. As far as we know, no transretinal measurements of oxygen levels have been performed after recanalized BRVO. Nevertheless, oxygen levels within the retina are not the only parameter that affects the retinal metabolism during vein occlusion. Other factors, including capillary hyperpressure, increased extracellular pressure, intraretinal hemorrhages, leukostasis, ²³ serous retinal detachment, secondary thrombosis and ischemia-reperfusion may also contribute to impair retinal and/or endothelial metabolism. In addition, it is not known whether retinal flow returns to normal in the previously occluded area after recanalization. The respective roles of these factors remain to be determined. 

The present findings may have important clinical implications for vein occlusion treatment. Vein decompression by separation of the artery and vein at the arteriovenous crossing has been advocated for treatment of BRVO, ²⁴ and optic neurotomy, for central retinal vein occlusion. ²⁵ However, it has not been clearly demonstrated that these procedures are followed by the repermeabilization of occluded capillaries. ²⁶ Our results suggest that even a short period of occlusion can prevent complete regression of capillary closure, and therefore that surgical benefits will be limited unless surgery is performed very early. 

It must be stressed that the experimental model of sudden, complete, and transient occlusion described herein contrasts with the progressive and often irreversible occlusion encountered in humans. Also, BRVO mostly occurs in elderly hypertensive patients who are likely to have an underlying endothelial arterial and/or venous disease. Animals with chronic focal stenosis of a retinal vein, for instance, by surgical placement of venous clips that reduce the venous diameter, should be a better model. However, our model is reproducible and inexpensive and constitutes an interesting approach to the study of the effects of acute vein obstruction and recanalization. Also, the recanalization of an occluded vein, which is unlikely to occur spontaneously in branch vein occlusion, is more frequent during central vein occlusion and thus this model may also contribute to the understanding of the sequelae of transient central retinal vein occlusion. Another advantage is that it can be used as a model of ischemia-reperfusion. 

In summary, transient occlusion of a retinal vein in rats leads to short- and long-term microvascular remodeling upstream of the occlusion site. Within minutes of vein occlusion, collateral circulation arises from the dilation of deep-layer venules. Early capillary dropout affects the three microvessel layers differentially, the intermediary microvessel layer being the most severely affected. We propose a model of intraretinal distribution of blood flow, which may account for the topography of these early microvascular changes. In the long term, these changes regressed incompletely with the recanalization of the occluded vein. Therefore, microvascular changes may become relatively independent of blood flow after a short period of occlusion. Despite the intrinsic limitations of the applicability of this model to human disease, our results suggest that even if therapeutic decompression of an occluded vein is performed early, it may not completely reverse capillary dropout. 

 Submitted for publication July 19, 2003; revised October 8, 2003; accepted October 10, 2003.

 Disclosure: O. Genevois, None; M. Paques, None; M. Simonutti, None; R. Sercombe, None; J. Seylaz, None; A. Gaudric, None; J.-P. Brouland, None; J. Sahel, None; E. Vicaut, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Successive confocal microscopy z-scans of retinal microvessels in the normal rat from the inner to the outer retina. (A) Superficial microvessel layer; (B) vertical capillaries; (C) intermediary layer; (D) deep layer. The vertical capillaries underlying the intermediary layer are not shown. (A, arrow) Direct connection of a superficial capillary to the vein. (E, F) Confocal scanning laser ophthalmoscopy (cSLO) frames focused on the superficial layer (E) and deep layer (F) illustrate the similarities between the in vivo and ex vivo images. The intermediary layer cannot, however, be imaged by cSLO. Scale bar, 200 μm.

Figure 2.

 

High-magnification confocal microscopy images of retinal microvessels in the normal rat. (A–E) Successive z-scans from the inner to the outer retina. (A, C, E) Superficial, intermediary and deep layer. (B, D) The vertical capillaries joining the three layers are visible. (A, arrowhead) A capillary connected to a major vein. (B, E, arrows) A vertical-connecting venule running between the deep and superficial layers. Scale bar, 50 μm.

Figure 3.

 

cSLO fluorescein angiogram immediately after vein occlusion, focused on the deep layer. Note the tortuosity of venules and leakage of dye into the deep layer around them.

Figure 4.

 

Illustration of collateral vessel formation after BRVO. cSLO fluorescein angiography centered on the artery bordering on the occluded territory, 2 days after vein occlusion.

Figure 5.

 

Vein occlusion site on studied days. Fluorescein angiography with confocal scanning laser ophthalmoscopy. Note the discrete residual vein stenosis at days 3 and 8, which disappeared by day 30.

Figure 6.

 

cSLO fluorescein angiography 30 days after vein occlusion, showing a major vein next to the occluded vein. The venous occlusion territory is on the right side of the figure. Despite recanalization of the neighboring occluded vein, the tortuosity of deep-layer venules (arrows) persisted on the side of the vein occlusion, whereas the venules on the opposite side were normal.

Figure 7.

 

Capillary density and microvessel diameter (arbitrary units; mean ± SD) upstream of the vein occlusion site (n = 6 at each time-point, except for control subjects, for which n = 16). Because results were similar in sham-treated and control eyes (data not shown), they were pooled to form the control group (C). Time points are days after laser treatment. *P < 0.01; **P < 0.0001, vs. controls; Mann-Whitney test.

Figure 8.

 

Illustration of venule dilation and capillary dropout after BRVO. Confocal microscopy examination of the three microvessel layers. Note the capillary dropout in the intermediary layer and the dilation of deep layer venules at day 1, which regressed incompletely by day 30.

Figure 9.

 

Retinal histology upstream of the venous occlusion site at day 30 (hematein-eosin stain). No morphologic changes were found in the affected retinal area.

Figure 10.

 

Schematic representation of the distribution of blood flow in the retina in the normal state (top) and after vein occlusion (bottom). The vitreous side is up. Dotted line in the inner retina represents the connections between superficial layer capillaries and major veins. After vein occlusion preferential routing of flow from superficial arterioles to the neighboring vein results in functional closure of intermediary layer capillaries. A, artery; V, vein.

The authors thank Alcon and Quentel Medical Corporations for providing the laser for the duration of the study.