Perivascular accumulation of pigment is associated with decreased vascular leakage in eyes with MacTel

To investigate the impact of perivascular pigment accumulation on vessel leakage and proliferation in MacTel, we compared the longitudinal courses of eyes with and eyes without pigment plaque de novo formation. A total of 1216 eyes from 608 patients of 12 study centers were evaluated. 69 eyes from 69 patients (mean age 61.9 [range 53–71] years; 37 [54%] females) were included and reviewed over a mean period of 41.6 months (range 24–60). 35 eyes (51%) showed a de novo development of pigmentary changes, and 34 eyes did not. Pigment plaques predominantly accumulated along vessels within the temporal parafovea (ETDRS subfield 5), usually sparing the fovea (Online Resource 1). Rarely, an extension of changes to the superior, inferior, or nasal parafovea (ETDRS subfields 2, 4, 3) was observed.

Longitudinal courses of eyes developing pigmented lesions differed from those without. A decrease in fluorescein leakage and stabilization of vessel proliferation was noted in all but one eye with pigment plaques. The observed effects were, however, focal and limited to vessels covered with pigment. In these eyes, vessels lacking pigment plaques showed stable, or rarely, increased leakage (observed in the nasal parafovea (ETDRS subfield 3) of 4/35 eyes; see Fig. 1). Notably, coverage of vessels with pigment was associated with a decrease in fluorescein leakage both in the early and late phase of fundus fluorescein angiography (FFA), suggesting a sealing, rather than a mere shadowing effect associated with perivascular pigment accumulation. In eyes without pigmentary changes, an increase in vascular leakage (in 16/34 eyes [47%]) or stable leakage (in 18/34 eyes [53%]; Table 1) was observed. Proliferation of vessels and increase in leakage were primarily observed within the temporal parafovea (ETDRS subfield 5). On optical coherence tomography (OCT), the de novo development of pigment plaques (see Fig. 2, left panel, cases 1–3) was associated with overall stable findings. Compared to the baseline visit, an increase in intraretinal hyper-reflectivity as well as increased shadowing of underlying structures within areas showing pigment plaques formation could, however, commonly be observed. Whether the latter signals on OCT were associated with perivascular pigment accumulation, vascular changes or a combination of both events can, however, not be determined, as discussed in previous imaging studies [16, 20, 21]. Exemplary longitudinal courses are illustrated in Figs. 1 and 2.

Fig. 1

Perivascular accumulation of pigment plaques decreases vascular leakage in patients with MacTel. Longitudinal courses of three exemplary eyes showing a de novo formation of perivascular pigment plaques, imaged with color fundus photography (CFP), blue-light autofluorescence (BAF), and fundus fluorescein angiography (FFA; early to intermediate phase and late phase). A decrease in fluorescein leakage can be observed in vessels covered with pigment (yellow borderline), while vessels lacking pigment may show an increase in leakage and proliferation (blue borderline). The right column illustrates enlarged BAF and FFA images of case #2 within the temporal and nasal parafovea, respectively

Fig. 2

Perivascular pigment accumulation stabilizes vessel growth and decreases neovascular exudation in eyes with MacTel. Longitudinal courses of exemplary eyes with (cases 1–3) and without (cases 4–6) a de novo development of pigment plaques. Perivascular pigment accumulation is associated with a decrease in fluorescein leakage on fundus fluorescein angiography (FFA), and overall stable findings on spectral domain-optical coherence tomography (SD-OCT). Note the increase in intraretinal hyper-reflectivity (black arrowheads) and shadowing of underlying structures (white arrowheads, OCT-scans, left panel) associated with the de novo development of pigment plaques (solid white arrowheads, CFP and FFA images, left panel). Eyes without pigmentary changes show an increase in fluorescein leakage that may be associated with the de novo development of exudative neovascularization (solid white arrowheads, right panel). Hemorrhages may occur. White dashed horizontal lines indicate the position of OCT-B-scans on corresponding CFP images; vertical lines indicate the position of pigment plaques on CFP and OCT-scans, respectively. CFP: color fundus photography

Exudative subretinal neovascularization is considered a severe, vision-threatening complication of MacTel, and is associated with severe vascular leakage. The de novo formation of exudative subretinal neovascular membranes was less frequently observed in eyes with, compared to eyes without, pigment plaques (in 1/35 [3%] eyes vs. 7/34 [21%]; Fisher’s exact test, p = 0.0275). Table 1 gives an overview of the clinical findings in this study cohort.

Table 1 Clinical findings in the study cohort of patients with macular telangiectasia type 2 (MacTel)

In summary, patients with MacTel showed (I) an accumulation of pigment plaques along abnormal retinal and subretinal vessels, (II) a decrease in vascular leakage that was associated with the development of pigment plaques, and (III) a decrease in exudative subretinal neovascularization associated with the presence of pigment plaques. Based on our findings in patients with MacTel, we hypothesized that: (I) Proliferating vessels may trigger the proliferation and perivascular accumulation of pigment; (II) Pigment plaques may be formed by RPE cells that undergo EMT, proliferate, migrate and accumulate along proliferating vessels; and (III) Perivascular pigment plaques may decrease vascular leakage and stabilize vessel proliferation, thus having a beneficial effect on the diseased retina.

To test these hypotheses, and to further evaluate disease mechanisms leading to perivascular pigment accumulation, we studied related changes in the Vldlr−/− mouse model. Similar to eyes with MacTel, the Vldlr−/− mouse model shows a proliferation of retinal vessels, formation of retinal-choroidal anastomoses and subretinal neovascularization. With disease progression, RPE-cells proliferate and accumulate along subretinal neovessels, and subsequently migrate along retinal vessels into the neuroretina [17,18,19].

Proliferating retinal vessels trigger perivascular pigment accumulation

We first set out to investigate whether vascular proliferation triggers the proliferation and perivascular accumulation of pigment. In the Vldlr−/− mouse model, retinal vessels begin proliferating around P12, followed by the growth of retinal vessels to the outer retina, and the formation of subretinal neovascular complexes around P16-P21 [22]. RPE-cells start proliferating around 4 weeks of age, accumulate along neovessels in the subretinal space, and subsequently migrate along retinal vessels into the neuroretina [18, 19, 23]. By inhibiting vascular proliferation using neutralizing antibodies against vascular endothelial growth factor (VEGF), we found a reduction in neovascular tuft formation. The ratio of pigmented to non-pigmented neovessels was, however, unchanged, and pigment plaques only developed along proliferating neovessels (see Fig. 3a), suggesting that neovessels precede, and are required for, pigment plaque formation.

Fig. 3

The Vldlr knockout mouse model mirrors vascular and pigmentary changes observed in MacTel. a: Proliferating vessels trigger perivascular pigment accumulation in Vldlr−/− mice. Neovascular (NV) tufts and perivascular pigment accumulation were analyzed using GS-lectin staining and bright field images in P28 Vldlr−/− mice treated with intravitreal injections of anti-VEGF (n = 6) or control IgG (n = 4) at P12. While anti-VEGF treatment significantly decreased the total number of NV-tufts (Mann-Whitney test, *p = 0.019), the ratio of pigmented to non-pigmented NV-tufts was unchanged. Pigment plaques only developed along proliferating vessels. Error bars indicate the median and interquartile range. b: Regulation of different genes coding for key molecules and inducers of epithelial-mesenchymal transition (EMT) in the retina and RPE of Vldlr−/− mice at P42. Changes in genes between Vldlr–/– and Vldlr−/+ mice, as analyzed using a PCR array for EMT, are shown. P values were calculated based on a Mann-Whitney test of the replicate 2(-Delta Ct) values for each gene in the Vldlr−/− and Vldlr−/+ groups. *P < 0.05, **P < 0.01 (n = 5 each). Median and IQR are shown for each gene. c: Dextran-angiography in a 12-months-old Vldlr−/− mouse depicts vessel and pigment proliferation and migration of pigmented cells along retinal vessels. Vessels covered with pigment show reduced dextran leakage (enlarged images, yellow arrowheads) compared to vessels not covered with pigment (white arrowheads). d: Pigment plaques cover proliferating vessels (stained with GS-lectin) and express ZO-1, indicating the formation of tight junctions along subretinal and retinal vessels in eyes of 10-months-old Vldlr−/− mice. e: Vldlr knockout (Vldlr−/−) mice show hyper-reflective changes at the level of the outer retina / retinal pigment epithelium (RPE; yellow and black arrowheads, respectively) on optical coherence tomography that resemble alterations observed in MacTel

Pigment plaques are formed by RPE cells in Vldlr−/− retinas and express similar markers as observed in eyes with MacTel

Previous findings in postmortem retinal samples of eyes with MacTel or retinitis pigmentosa indicated that intraretinal pigment plaques originated from the RPE [2]. Intraretinal lesions were found to express the epithelial cell marker cytokeratin18 (CK18), that is specific to RPE-cells in the retina, but were negative for RPE65. Markers for mesenchymal cells (alpha-smooth muscle actin [ASMA]) and macrophages/ microglia (IBA1) were also evaluated, but found to be absent [2]. To verify these findings in the Vldlr−/− mouse model, we evaluated similar markers as previously described [2]. Proliferating RPE cells within the subretinal space expressed CK18 and RPE65. Some, but not all of these cells also showed immunoreactivity for ASMA (see Online Resource 2), indicating EMT of the RPE. Intraretinal pigment plaques, on the other hand, expressed CK18, but neither RPE65 nor IBA1 were detected (see Online Resources 2 and 3). The expression of ASMA was observed in single intraretinal pigmented lesions. The latter were, however, overall smaller and less dense compared to lesions lacking ASMA expression, indicating a transitional, possibly less mature, stage of these lesions (Online Resources 2 and 3).

RPE-cells undergo EMT in Vldlr−/− retinas

Under physiologic conditions, the RPE is formed by a monolayer of polarized cells. Disintegration of the RPE monolayer and proliferation and migration of RPE-cells have been described in several degenerative retinal diseases, and have been attributed to RPE cells transitioning from an epithelial to a mesenchymal state [5]. To test whether RPE cells underwent EMT in the Vldlr−/− mouse model, we compared gene expression levels of known EMT-related genes in the RPE of Vldlr–/– mice and control Vldlr−/+ littermates at P42. At this timepoint, RPE cells have been shown to proliferate and accumulate along subretinal neovessels and start migrating along retinal vessels into the neuroretina [17,18,19]. Using qPCR arrays, we found an enrichment of genes coding for EMT pathways (SNAIL1/2) and different mesenchymal markers (vimentin, fibronectin, N-cadherin) in the RPE of Vldlr-/- mice. Genes coding for epithelial markers (beta-catenin, E-cadherin, zonula occludens-1 [ZO-1]), on the other hand, were decreased (Fig. 3b), indicating that RPE cells underwent EMT in this model.

Next, we tested mRNA expression levels of known inducers of EMT in the retinas and RPE of Vldlr−/− mice. The largest differences between Vldlr–/– and heterozygous control littermates were found in fibroblast growth factor-2 (FGF2), which was increased 4-fold, and in tumor-necrosis factor-alpha (TNFA), which was increased 2.5-fold. FGF2 is a known driver of EMT that, among other factors, has been described to play a role in inducing EMT in RPE cells in proliferative vitreoretinopathy (PVR) [24]. FGF2 is also known to play a role in inducing subretinal fibrosis, and has been shown to have pro-angiogenic properties [25, 26]. TNFA is a proinflammatory cytokine and a known inducer of EMT in RPE cells [5, 27]. Elevated levels of TNFA have been detected in vitreous samples and epiretinal membranes of patients with PVR [27,28,29]. In vitro, TNFA has been shown to induce RPE cells to upregulate EMT markers and mesenchymal key molecules [30]. Increased expression levels of TNF have previously been found in the retinas of Vldlr−/− mice, and in particular, at the level of the deep retinal plexus [22].

Perivascular pigment decreases neovascular leakage and proliferation in Vldlr−/− retinas

Similar to eyes with MacTel, we found that in the Vldlr−/− mouse model vessels covered with pigment showed reduced dextran leakage compared to vessels not covered with pigment (Fig. 3c). Perivascular pigment plaques expressed zonula occludens-1 (ZO-1), indicating the formation of tight junctions around proliferating vessels, thereby possibly reducing vascular leakage (Fig. 3d). Furthermore, on OCT, Vldlr−/− mice showed hyper-reflective changes at the level of the outer retina/ RPE that resemble alterations observed in MacTel (Fig. 3e). These changes have been proposed to represent outer retinal neovascularization and proliferating RPE-cells [16]. Next, we set out to investigate whether inhibiting EMT of the RPE may impact neovascular leakage and proliferation in the Vldlr–/– model. Mice treated with neutralizing antibodies against FGF2 or TNFA from P21 to P42 showed a significant increase in vascular leakage and in the size of neovascular complexes as well as a significant decrease in perivascular pigment accumulation at P42 in comparison to IgG-treated control animals (Fig. 4a-d). Vascular leakage and perivascular pigment accumulation showed a negative correlation in both treated and control animals (Fig. 4e).

To test whether the observed morphological changes were associated with the inhibition of EMT we compared gene expression levels of EMT-related genes in the RPE of Vldlr−/– mice treated with FGF2, TNFA or control IgG. While no changes were observed for EMT pathways, genes coding for epithelial markers were enriched, and mesenchymal markers were decreased in animals treated with FGF2 or TNFA, indicating the inhibition of EMT (Fig. 4f-g).

Fig. 4

Inhibition of epithelial-mesenchymal-transition (EMT) leads to enhanced neovascular proliferation and leak in the Vldlr−/− model. a-c: Dextran angiography and GS lectin staining in flat mounted (a, upper panel) or cryo-sectioned (a, middle and lower panel) Vldlr−/− mice treated with intravitreal injections of neutralizing antibodies against TNFA (n = 7) or FGF2 (n = 7) at P21 showed a significant increase in dextran leakage (a, upper panel; c), a significant increase in size of neovascular (NV) complexes (b), and a significant decrease in pigment accumulation (number of pixels positive for pigment; d) at P42 compared to IgG-treated controls (n = 7). Yellow-dotted boxes in a (upper panel) show enlarged neovascular complexes in treated eyesGS-lectin staining and bright field images of cryo-sectioned Vldlr−/− (a, middle and lower panel) illustrate decreased pigment proliferation and perivascular accumulation in eyes treated with anti-TNFA or anti-FGF2. e: Dextran-leakage (number of pixels positive for dextran) was negatively correlated (Pearson r = − 0.78; p < 0.0001) with pigment accumulation (number of pixels positive for pigment) in flat-mounted Vldlr−/− eyes. f and g: Changes in genes coding for key molecules of EMT in Vldlr–/– mice at P42 treated with either anti-TNFA (f) or anti-FGF2 (g) compared to IgG-treated Vldlr−/− mice (treatment at P21), as analyzed using a PCR array for EMT, are shown. While mesenchymal key molecules (white bars) were downregulated, epithelial key molecules were enriched (grey bars). P values were calculated based on a Mann-Whitney test of the replicate 2(-Delta Ct) values for each gene. *P < 0.05, **P < 0.01 (n = 4 each). GS-lectin staining and bright field images of cryo-sectioned Vldlr−/− (a, middle and lower panel) illustrate decreased pigment proliferation and perivascular accumulation in eyes treated with anti-TNFA or anti-FGF2. e: Dextran-leakage (number of pixels positive for dextran) was negatively correlated (Pearson r=-0.78; p < 0.0001) with pigment accumulation (number of pixels positive for pigment) in flat-mounted Vldlr−/−eyes. f and g: Changes in genes coding for key molecules of EMT in Vldlr–/–mice at P42 treated with either anti-TNFA (f) or anti-FGF2 (g)compared to IgG-treated Vldlr−/− mice (treatment at P21), as analyzed using a PCR array for EMT, are shown. While mesenchymal key molecules (white bars) were downregulated, epithelial key molecules were enriched (grey bars). P values were calculated based on a Mann-Whitney test of the replicate 2(-Delta Ct) values for each gene. *P < 0.05, **P< 0.01 (n= 4 each)

In summary, we suggest that the perivascular accumulation of RPE-cells may stabilize neovascular proliferation and leakage, thereby exerting a beneficial, protective effect on the diseased retina. Figure 5 summarizes the herein proposed mechanisms in a schematic illustration.

Fig. 5

Schematic illustration of the proposed interplay between RPE-cells and retinal vessels in the Vldlr−/− model. a: Proliferating vessels of the deep retinal plexus grow to the outer retina, form neovascular tufts and come in contact with the RPE. RPE cells transition from an epithelial to a mesenchymal state, proliferate and migrate along retinal vessels into the neuroretina, forming perivascular plaques. b: Eyes treated with intravitreal antibodies against VEGF (anti-VEGF) show a reduction in vessel proliferation and decreased numbers of neovascular tufts, while EMT of the RPE is not impacted. c: In eyes treated with intravitreal antibodies against TNFa or FGF2, epithelial-mesenchymal transition (EMT) of the RPE is inhibited. RPE cells show reduced proliferation, perivascular accumulation and migration. Neovascular tufts show more leakage and increased lateral growth. Created with BioRender.com