Identification of Structures Labeled by Indocyanine Green in the Rat Choroid and Retina Can Guide Interpretation of Indocyanine Green Angiography

Abstract

Purpose: Indocyanine green (ICG) is an albumin and lipoprotein binding dye absorbing in the far red used in angiography to visualize choroidal vessels (ICG angiography [ICGA]). To guide interpretation, ICG transport in the choroid, RPE, and retina of rats was studied.

Methods: Two conditions were used: RPE/choroid organoculture, incubated for 45 minutes in DMEM medium, 1% fetal bovine serum containing 0.25 mg/mL ICG and RPE/choroid and neural retina flat-mounts at 1 and 6 hours after intravenous ICG injection. Early and late sequences of ICGA were recorded until 6 hours. Ultra-deep red confocal microscope was used to localize ICG in flat-mounts and immunohistochemistry was performed for caveolin-1, tryptase (mast cell marker), and tubulin β3 (a nerve marker).

Results: In the organoculture, ICG penetrated homogeneously in the cytoplasm and stained the membranes of the RPE. At 1 hour after intravenous injection, ICG appeared in fine granules in RPE, partly labeled with caveolin-1 and decreasing at 6 hours. At 1 hour and 6 hours, ICG was found in the retinal vessels, faintly in the inner retina, and in the photoreceptor outer segments at 6 hours. In the choroid, ICG colocalized with mast cells, immunostained with tryptase, and accumulated along the large tubulin β3–labeled nerve bundles. The hypothesis was raised on the interpretation of late ICGA infrared photography in case of transthyretin amyloidosis with neuropathy.

Conclusions: Beside being a vascular dye, ICG is transported from the vessels to the RPE toward the outer retina. It stains mast cells and large choroidal nerves. These observations could help the analysis of ICGA images.

Indocyanine green angiography (ICGA) has been used in clinical practice for more than 50 years, although its indication has decreased with the wide use of spectral domain optical coherence tomography and optical coherence tomography angiography¹ because the main indication for ICGA is the visualization of choroidal neovascularization, often well-identified by noninvasive imaging methods. Indeed, ICG is an amphiphilic, water soluble, tricarbocyanine dye with a molecular mass of approximately 750 Da, which is >95% bound to high-density lipoprotein (HDL) and low-density lipoprotein (LDL)² and to albumin after intravenous (IV) injection.³ It is excited by wavelengths between 750 and 800 nm and emits fluorescence that peaks at approximately 830 nm, avoiding interference with autofluorescence from surrounding tissues. Because of these chemical properties and its binding to plasma proteins and lipoproteins, ICG remains in normal choroidal vessels, including the choriocapillaris, and can be visualized through the pigmented RPE, making it an ideal dye to image the choroidal vasculature with high contrast and sensitivity. 

The clinical significance of ICGA is mostly based on the spatiotemporal localization of the fluorescence emitted by ICG after IV injection. The angiographic sequence allows to follow the very rapid filling of the arteries and choriocapillaris, followed by filling of the choroidal veins and drainage through the vorticose veins. Appearance of hyperfluorescence reflects leakage from choroidal vessels, whereas hypofluorescence indicates decreased perfusion or masking by absorbing structures, but the kinetics of the hyperfluorescence or hypofluorescence, and their localization differs depending on the pathology and the disease mechanisms. We will not detail here the ICGA semiology that has been well-reviewed by others who showed its value for diagnosis and even for prognosis in some chorioretinal diseases.^1,4–6 When interpretating ICGA, beside its kinetics, several factors that influence ICG fluorescence should be considered, including its concentration and its binding to different surfaces. Particularly, the fluorescence intensity might increase at low ICG concentrations (1 µM), but strikingly decrease at high concentrations (10 µM).⁷ Binding to albumin with high affinity might also increases ICG fluorescence.⁷ This property was used to detect proteins in capillary electrophoresis because ICG is only weakly fluorescent in dilute aqueous solution, but its near infrared fluorescence is highly enhanced when it binds to proteins and specifically albumin.⁸ More recently, Jang et al. used plasmon resonance and saturation binding assay to show that ICG fluorescence in vivo is higher for albumin-bound ICG (2.1-fold at 1 hour after injection).⁹ 

Beside visualization of the choroidal vasculature, ICGA has the potential to provide additional indirect information on the metabolic capabilities of the RPE and on the intracellular trafficking of albumin and lipoproteins in endothelial and in RPE cells.¹⁰ Chang et al. demonstrated that ICG is slowly internalized by RPE cells in vivo in the primate and in vitro in human RPE cells.^11,12 Various vesicular transports have been described in RPE cells for HDLs thar are internalized after binding to scavenger receptor class B type I, which is a mechanism to import xantophylls into the retina¹³ and, for albumin and LDL, through caveolin-mediated transcytosis, because caveolin-1 is highly expressed in RPE cells.¹⁴ The vesicular transports could explain the kinetics of ICG internalization in RPE cells.^11,12 In endothelial cells, HDL partially colocalized with LDL, albumin, and transferrin in intracellular vesicles,¹⁵ suggesting that ICG could also be internalized in endothelial cells through vesicular transporters. In other fields of medicine, ICG is used as an adjuvant to surgery to localize and preserve nerves and to perform lymphography,^16,17 because ICG labels nerves,¹⁸ and this network is very dense in the human choroid.¹⁹ However, whether ICG could label choroidal nerves in normal or pathologic conditions has not yet been questioned. 

Understanding the in vivo kinetics of ICG fluorescence requires considering the vascular and RPE macromolecular transport mechanisms and how they might be impaired in pathogenic conditions, as well as taking into account the specificities of the dye, which fluorescence depends on its concentration, binding capacities, and pH, which are all susceptible to changes in the different studied diseases. 

A more extensive identification of the fate of injected ICG and its interaction with the retina and choroid could guide our interpretation of the clinical images. With this objective in mind, we performed experiments to analyze how ICG is distributed in the retina and choroid in rats, either after IV injection or after ex vivo organoculture. 

Methods

Animals

Animals were kept in pathogen-free conditions with food, water and litter, and housed in a 12-hour/12-hour light/dark cycle. Sprague Dawley albinos male and female rats between 3 and 12 months old were used (n = 22). Albinos rat were used to allow a better visualization of choroidal immune-labeled structures. 

ICGA

Rats were anesthetized by IP injection of 100 mg/kg of ketamine (Clorkétam 1000, Virbac France) and 4 mg/kg of xylazine (Rompun 2%, Bayer Healthcare, Loos, France). After pupil dilation, ICG (200 µL, 2.5 mg/mL INFRACYANINE, SERB, Paris, France) was injected intravenously in the tail of rats. ICGA was performed using Heidelberg Retina Angiograph II (Heidelberg Engineering, Inc., Dossenheim, Germany) to image choroidal circulation and visualize ICG tissue staining. Pictures were recorded at 1 to 3 minutes, 10 to 15 minutes, 30 minutes, 1 hour, and 6 hours. Rats were euthanized by Euthasol-VET (300 mg/kg, Dechra, Northwich, UK) either at 1 hour (n = 6) or 6 hours (n = 4). After enucleation, eyes were fixed in 4% paraformaldehyde for 15 minutes at room temperature. After washing with DPBS (Thermo Fisher Scientific, Illkirch-graffenstaden, France), eyes were dissected, the RPE–choroid–sclera complex and the neural retina were postfixed with acetone at –20°C for 10 minutes and proceeded for flat mounting. 

Organoculture

Rats (n = 9, 18 eyes) were euthanized by Euthasol-VET. After enucleation and dissection of the anterior segment, the posterior segment of the eyeball including retina, choroid and sclera was incubated immediately for 45 minutes in DMEM medium (41965039, Thermo Fisher Scientific, Illkirch-graffenstaden, France) with 1% fetal bovine serum (10270106, Thermo Fisher Scientific) and 10% ICG (2.5 mg/mL, Infracyanine) at 37°C (5% CO₂). After washing with DPBS, the neuroretina was removed. The RPE–choroid–sclera complex was fixed with 4% paraformaldehyde, then post-fixed with acetone for 10 minutes at −20°, and proceeded to flat mounting. We chose a 10-fold higher concentration for the ex vivo experiments as compared with the in vivo experiments, because this experiment was designed as a positive control for RPE internalization of ICG. 

Flat Mounting and Immunofluorescence

Four radical incisions were performed on the neuroretina and RPE–choroid–sclera complex. To observe ICG staining, tissues were directly flat mounted with Dako Omnis Fluorescence Mounting Medium (Agilent, Les Ulis, France) and counter stained with DAPI (1:5000) for nuclei staining. Ultra-deep red confocal microscope (Leica LAS X software, STELLARIS 5, Leica Microsystems, Wetzlar, Germany) with an excitation wavelength at 700 nm was used to visualize and capture ICG staining. The RPE–choroid–sclera complex from three control rats without ICG injection was used to check the autofluorescence at this wavelength. 

The RPE–choroid–sclera and neural retina were also used for immunohistochemistry. Tissues were permeabilized with 0.01% triton X100 (Merck, Darmstadt, Germany) in DPBS, blocked with 10% normal goat serum (G6767, Merck) in DPBS, and then incubated with primary antibodies (Table) at appropriate dilution for 7 days at 4°C under gentle agitation. After washing with 0.01% Triton X100/DPBS, tissues were incubated with adequate secondary antibodies respectively: AlexaFluor 488 goat anti-rabbit IgG (1:500, Thermo Fisher Scientific) and AlexaFluor 488 donkey anti-mouse IgG (1:500, Thermo Fisher Scientific). The nuclei were counterstained with DAPI (1:5000). RPE–choroid–sclera and neuroretina were flat mounted with Dako Omnis Fluorescence Mounting Medium and observed with confocal microscope. Negative controls omitting the primary antibody were tested according to the above protocol and did not show nonspecific labelling at the level of the RPE or at the level of the choroid. 

Table.

 

List of primary antibodies

 

Case Report of Amyloidosis

A 52-year-old woman with hereditary transthyretin amyloidosis caused by the p.Val50Met variant was referred for an ophthalmological assessment. She had undergone a liver transplant 16 years earlier and was experiencing severe sensorimotor and autonomic neuropathy. The patient had a history of amyloidosis affecting both eyes, including vitreous opacities and severe glaucoma that required two filtering surgeries and phacovitrectomy in the left eye. Her best corrected visual acuity was 20/25 in the right eye and 20/200 in the left eye, with the IOP measured at 30 mm Hg in the right eye and 10 mm Hg in the left eye. Biomicroscopic examination disclosed bilateral typical fringed pupils, along with deposits on the anterior lens capsule in the right eye and a posterior chamber intraocular lens in the left eye. Fundus examination revealed mild vitreous deposits in the right eye and a clear vitreous cavity with amyloid remnants in the peripheral vitreous of the left eye (which had undergone vitrectomy), as well as bilateral diffuse vascular sheathing and excavated optic discs. 

Results

ICG Is Internalized in RPE Cells Through Vesicular Transport

No fluorescent signal was observed when RPE/choroid flat mounts were excited at 700 nm, showing no autofluorescence of the tissues at this wavelength (Fig. 1A). This negative control was analyzed systematically in each experiment. The fluorescence of ICG allowed to observe it was internalized homogeneously into RPE cells and was also located at the cell membrane at one hour after posterior segment incubation in the medium containing ICG (0.25 mg/mL of DMEM 1% fetal bovine serum, final concentration) (Fig. 1B). One hour after IV injection of ICG (0.5 mg), the fluorescent signal was located inside RPE cells, mostly concentrated into vesicles (Fig. 1C, inset, arrows). At 6 hours after IV injection, intracellular ICG-labeled vesicles were still visible although at lower density (Figs. 1D, 1E, inset, white arrows). 

Figure 1.

 

Deep infrared confocal imaging of the RPE/choroid flat mounts. (A) No fluorescence when imaging is performed from the apical side without ICG. (B) At 45 minutes after incubation in ICG, homogenous fluorescence of RPE cells and of their membranes (inset). (C) One hour after ICG IV injection, granular hyperfluorescence of the RPE (inset). (D) Six hours after ICG IV, the granular hyperfluorescence decreases. (E) Image from D with nuclei DAPI staining.

Immunolocalization of caveolin-1 on RPE/choroid flat mounted at one hour after IV of ICG (Figs. 2A, 2B), showed that at least part of the ICG-labelled intracellular vesicles were stained with caveolin-1 (Fig. 2C, inset white arrows), demonstrating that ICG is partially transported in RPE through caveola-mediated transcytosis. Specificity of the antibody is shown by the absence of staining on negative controls, omitting the primary antibody (Figs. 3A, 3D). 

Figure 2.

 

Deep infrared confocal imaging of RPE flat mounts and caveolin-1 immunohistochemistry at 1 hour after IV injection of ICG. (A) Deep infrared imaging of RPE showing granular staining with nuclei DAPI staining. Inset shows magnification. (B) Caveolin-1 immunohistochemistry on RPE flat mount shows vesicular staining at the membrane and within the cytoplasm, with nuclei DAPI staining. Inset shows magnification. (C) Merge of images A and B shows partial colabelling of ICG with caveolin-1 with nuclei DAPI staining. Inset shows magnification with yellow costained vesicles (white arrows), caveolin-1 vesicles (green arrows) and ICG-stained vesicles (red arrows). Scale bar, 20 µm.

Figure 3.

 

Negative controls without primary antibody in RPE/choroid flat mounts. (A, D) AlexaFluor 488 goat anti-rabbit in the RPE, negative control for Caveolin-1. (B, E) AlexaFluor 488 goat anti-rabbit in the choroid, negative control for ionized calcium-binding adapter molecule 1. (C, F) AlexaFluor 488 donkey anti-mouse in the choroid, negative control for tryptase and tubulin β3.

ICG Is Transferred From the Choroid and RPE Toward the Outer Retina

At 1 hour, ICG was still located in retinal vessels, probably inside endothelial cells, with a faint ICG signal located around retinal vessels (Figs. 4C, 4D), but no signal was observed either in the outer segments or in the outer plexiform layer (Figs. 4A, 4B, and 5). At 6 hours, the ICG signal was present in the photoreceptor outer segments (Fig. 4E and 5), with a faint signal at the level of the outer plexiform layer and a decreased ICG signal in the retinal vessels (Figs. 4G, 4H). Cross-sections of the flat-mounted neural retina images confirmed the presence of ICG in the outer segments of the photoreceptors at 6 hours and increased staining from 1 to 6 hours in the ganglion cell layer (Figs. 5A, 5B). For better visualization of ICG distribution in the retinal layers, videos of the full Z stacks confocal fluorescence imaging of neural retina flat mounts are available as Video 1 (1 hour) and Video 2 (6 hours). 

Figure 4.

 

Deep infrared confocal imaging of neural retina flat mounts. (A–D) At 1 hour after ICG IV injection: outer segments (A), outer plexiform layer (B), ganglion cell layer (C), and ganglion cell layer with DAPI nuclei staining (D). (E–B) AT 6 hours after ICG IV injection: outer segments (E), outer plexiform layer (F), ganglion cell layer (G), ganglion cell layer with DAPI nuclei staining (H). Scale bar, 50 µm.

Figure 5.

 

Cross-sections of flat-mounted neural retina confocal images. Transverse pictures showing ICG with DAPI nuclei staining, from photoreceptor outer segments to the ganglion cell layer at 1 hour (A) and 6 hours (B) after ICG IV injection.

 

Video 1.

 

Video of Z-stack confocal fluorescent images of neural retina flat mounts at 1 hour after ICG IV injection. ICG with DAPI nuclei staining, from photoreceptor outer segments to the ganglion cell layer. Microscope: STELLARIS 5 (Leica Microsystems). Magnification: 40, distance: 73 µm, pixel size: 1 µm. Excitation wavelength: DAPI, 405 nm; ICG, 700 nm. Movies are available on the journal website.

 

Video 2.

 

Video of Z-stack confocal fluorescent images of neural retina flat mounts at 6 hours after ICG IV injection. ICG with DAPI nuclei staining, from photoreceptor outer segments to the ganglion cell layer. Microscope: STELLARIS 5 (Leica Microsystems). Magnification: 40, distance: 78 µm, pixel size: 1 µm. Excitation wavelength: DAPI, 405 nm; ICG, 700 nm. Movies are available on the journal website.

ICG Stains Mast Cells and Allows to Identify Large Choroidal Nerves

When imaging the choroid from the scleral side, we could identify round large cells (15–30 µm in diameter) positively stained by ICG, aligned along the vessels, both after direct incubation of the posterior segment in ICG and 1 hour after IV injection (Figs. 6A, 6B). Higher magnification showed intracellular granules positively stained with ICG (Fig. 6C). Immunostaining with an antibody that recognizes tryptase (a marker of at least part of the mast cells)²⁰ confirmed that those cells were mast cells (Figs. 6D–F) and not monocytic or dendritic cells, since they are not stained with ionized calcium-binding adapter molecule 1 (Fig. 6G). Negative controls are shown in Figures 3B and 3E for ionized calcium-binding adapter molecule 1 and in Figures 3C and 3F for tryptase. 

Figure 6.

 

ICG staining of mast cells. (A) ICG staining of perivascular cells 45 minutes after incubation of the RPE/choroid in ICG. (B, C) ICG staining of perivascular cells one hour after IV ICG injection with intracellular ICG positive granules (C). (D–F) Colabeling of ICG-stained cells (D) with tryptase (E) showing colabeled cells (F) (arrows). (G) Ionized calcium-binding adapter molecule 1 (IBA1) immunohistochemistry shows that ICG-positive cells are not labelled by IBA1.

The optic nerve head was only stained with ICG after incubation (Fig. 7A) and not after IV. However, both after incubation and at 1 hour after IV of ICG, we identified filamentous ICG-positive bundles (Figs. 7A, 7B) resembling choroidal nerves. Immunostaining with beta-tubulin 3, a nerve marker,²¹ indicated that ICG binds the nerve bundle (Fig. 7E, inset white arrows), although some nerve fibers seemed to be also stained with ICG (Fig. 7H, inset dark star). The negative control for the antibody is shown in Figures 3C and 3F. 

Figure 7.

 

Large choroidal nerves identification by ICG. (A) Deep infrared of choroid flat-mount at 45 minutes after ICG incubation showing optic nerve head and large nerve. (B, C, F) Deep infrared of choroid at 1 hour after ICG IV injection showing that ICG-stains large nerves. (D, G) Tubulin β3 (TUBB3) immunohistochemistry showing nerves fibers. (E, H) Costaining of ICG with TUBB33 showing ICG around nerve fibers (E) (inset white arrow).

Interpretation of ICGA in the Rat

At the very early phase of ICGA (≤1 min), the long ciliary posterior vessels emerging near the optic nerve are filled with ICG with a very rapid filling of the large veins, which can be followed in the midperiphery (Fig. 8A, 30 seconds, top and bottom). Large retinal vessels are filled (Fig. 9A) and vorticose veins are progressively visible (Fig. 9A). 

Figure 8.

 

ICGA in albinos rats up to 1 hour. ICGA of Sprague Dawley albinos rats, showing early (30 seconds) labelling of the long posterior ciliary vessels (LPCV) emerging at the posterior pole inferiorly to the optic nerve (A, top), and continuing toward the mid periphery (A, bottom). Optic nerve hyperfluorescence is seen in the very early phase with a peripapillary hyperfluorescent ring (A, top), whose intensity is attenuated, though still present, at 10 minutes and 1 hour (B, top, and C, top, respectively). At 10 minutes, most of the retinal and choroidal vessels are visualized as hypofluorescent structures against a homogeneous hyperfluorescent background (B, top and bottom). Along the hypofluorescent LPCV in the mid periphery, a linear hyperfluorescent signal can be observed (B, enlarged image, yellow arrow). One hour after ICG injection (C), the contrast between background fluorescence and vascular hypofluorescence becomes more evident. The granular background appearance in this late phase (C, top right and inset) could allow visualization of other, nonvascular components of the fundus. The midperiphery pictures at 1 hour show that the hyperfluorescence along the LPCV is maintained (C, bottom left and right, yellow arrow).

Figure 9.

 

ICGA in albinos rats up to 6 hours. ICGA of Sprague Dawley albinos rats, showing early (1-minute) retinal and choroidal vascularization and venous drainage with vorticose veins progressively visible (A, all images). At 10 minutes, a closer analysis of the peripapillary hyperfluorescence shows linear hyperfluorescent structures (B, top left and bottom enlarged left, yellow arrows). The choroidal hypofluorescent vascular component is clearly visible against a hyperfluorescent background (B, top right). In the midperiphery, long posterior ciliary vessels are surrounded by a linear hyperfluorescence (B, bottom right). Six hours after ICG injection, a background fluorescence is detected underneath the dark retinal and choroidal vessels (C, top left). The irregular peripapillary hyperfluorescence is showing detail (C, bottom left) where distinct hyperfluorescent structures are observed adjacent to vessels (white arrows). In the midperiphery (C, top right), irregular hyperfluorescence corresponds to linear perivascular structures (C, top and bottom right, yellow arrows).

At 10 minutes, ICG can barely be detected in the long posterior ciliary vessels and in retinal vessels, whereas a homogeneous background fluorescence can be detected with a hyperfluorescence around the optic nerve head (Fig. 8B, top, and Fig. 9B). A closer analysis of the peripapillary fluorescence shows hyperfluorescent tracks along the vessels (Fig. 9B, inset yellow arrows). At this time point, deep focalization probably shows bright fluorescence from the supra choroid and from the RPE. Along the long ciliary posterior vessels at the periphery, a hyperfluorescent signal can be detected, and could correspond with branches of the ciliary nerves that travel along the vessels (Figs. 8B and 9B, lower image yellow arrows). 

At 1 hour, homogeneous background fluorescence is observed with dark vessels being visible underneath (Fig. 8C). Higher magnification could detect possibly the RPE layer (Fig. 8C, inset). There is still hyperfluorescence around the optic nerve and along the large ciliary vessels that could correspond to nearby nerves (Fig. 8C, yellow arrow). 

At 6 hours, the low hyperfluorescent background is still visible (Fig. 9C), possibly corresponding with the RPE, but also with the outer segments with hyperfluoresence remaining around the optic nerve head with tracks (Fig. 8C, white arrows). Surrounding the choroidal vessels, hyperfluorescence could indicate choroidal nerves being labelled with ICG at this time point (Fig. 9C, yellow arrows). 

Interpretation of Late Phase ICGA in a Case of Amyloidosis

Amyloid deposits are visible on color fundus photographs as white-yellowish vascular sheathings (Fig. 10A, white arrowheads, magnification in the dark circle) that are not labelled by ICG at any time point and particularly not at the late phase of the ICGA (Figs. 10B, 10C). In contrast, on the late ICGA image (>20 minutes), typical ICG-positive elongated structures are visible in mid periphery and in the periphery of the fundus (Fig. 10C), previously identified as vascular amyloid deposits in choroidal vessels, mostly arteries.²² The reason why ICG would label the choroidal amyloid vascular deposits but not the retinal vessel amyloid deposits is unclear, and we hypothesize that ICG could stain the pathological nerves of these patients presenting neuropathy. Superimposition of the late ICGA image with color fundus and comparison with the ICG stained vessels at the early ICGA phase, do not show clear colocalization between the elongated structures and the choroidal vessels (Fig. 10, white circle on A, B, and C), although small arteries could be hard to identify. The B-scan optical coherence tomography, at the cross-section with one of the elongated ICG positive structure shows that the ICG-positive structure localizes with a round structure at the vicinity of a large choroidal vein, that does not contain the hyporeflective inside as observed in vessels. 

Figure 10.

 

Retina images of a 52-year-old woman with hereditary transthyretin amyloidosis and amyloid neuropathy. (A) Color fundus showing white yellowish vascular sheathings observed along retinal vessels (white arrowheads, dark circle with magnification). (B) At 2 minutes, ICGA infrared photography. (C) At 27 minutes, ICGA infrared photography, showing no staining of the amyloid deposits in retinal vessels (white arrowheads) and the ICG stained elongated dots and tracks. (A–C, white circle) The same area with superimposition of C and D on the magnified image in C. (D) B-scan along the ICGA stained elongated structures with the green line indicating the cross-section with one of them. Magnification shows that it does not correspond to the typical image of a choroidal vessel.

Discussion

We have used albinos rats to follow the fate of ICG after either IV injection or direct incubation of the ocular tissues from the posterior segment of the eye in ICG and we have imaged the fluorescence ocular tissues at late time points as compared with the classical angiographic sequence. The idea behind this experimental setting was to help understand clinical images taken at the late phases of the angiographic sequence, considering that the direct observation of ICG using confocal microscopy on flat-mounted tissues is much more sensitive than in vivo angiography. Indeed, the late phase of ICGA remains difficult to interpret in several clinical conditions. We recently analyzed the midphase hyperfluorescent plaques (6–10 minutes) observed on ICGA in patients with central serous chorioretinopathy and discussed the reason for the complete fading of fluorescence at later time point (>20 minutes), which differs from the hyperfluorescence linked to type 1 macular neovascularization in AMD that remains fluorescent on the late phases.^23,24 The reason for the intriguing ICG kinetics in multiple evanescent white dot syndrome was questioned recently by Gaudric and Mrejen.²⁵ 

The choice for albinos rat was made to allow immunolocalization of various structures together with ICG imaging, while decreasing the absorption due to melanin. Interpretation of our observation must thus consider this important difference with the pigmented human RPE/choroid. The use of a confocal microscope with simultaneous lines from 405 to 685 nm and detectors on three spectral channels provided high photon detection efficiency and extremely low dark noise, as well as extended detection into the near infrared of up to 850 nm. In addition, we could overlap infrared imaging and other fluorophores to perform immunolocalization. 

As previously described by several authors, we observed that either after ICG IV injection or after incubation of ocular tissues at 37°C, ICG-bound molecules were internalized into the RPE cells, although ICG accumulated also at the cell membrane after incubation and not after IV injection. After IV injection in nonhuman primates, Chang et al. showed that ICG was internalized within 15 to 50 minutes with increasing fluorescent signal in the RPE cells at the later time points.¹² Several studies were also performed in rodents showing not only that ICG was internalized by RPE cells after IV injection, but also that it remained detectable for an extended period of time and up to 28 days, particularly when a high dose of ICG (5 mg/kg) was injected.²⁶ In our experiments, we used 1.5 to 2.0 mg/kg, which corresponds with a high ICG dose, explaining that the ICG staining remained at 6 hours. The precise localization of ICG showed that at least part of it is transported in vesicles that were positively labeled with caveolin-1 antibodies, suggesting that caveolae-mediated transcytosis is involved in ICG transport, which is a well-known mechanism for albumin transport within cells²⁷ and a mechanism for albumin transcytosis in endothelial microvascular cells.²⁸ Transcytosis of LDL is also regulated by caveolin 1–mediated mechanisms in endothelial cells.²⁹ In the retina, caveolin-1 is expressed in retinal vascular cells, Müller glial cells, and RPE cells,³⁰ but the exact role of caveolin-1 in the protein and lipoprotein transports between the choroidal circulation and the neural retina is not yet fully understood. Lipoproteins, to which ICG is highly bound, are also transported from the choroidal blood flow toward the RPE and are responsible for the delivery of vitamin A, carotenoids, and lutein and zeaxanthin toward the inner retinal layers in the macula through the RPE,³⁰ demonstrating that transcytosis can occur also to allow the exit of molecules at the apical side of the RPE. Tserentsoodol et al. showed that, after IV injection of fluorescently labeled LDL and HDL, a fluorescent signal was observed in the choriocapillaris, the RPE, and part of the neural retina at 2 hours, and that the signal was detected in the outer segment of photoreceptors at 4 hours,³¹ which is consistent with our observation of an ICG signal in the outer segments at 6 hours and not at 1 hour after IV injection. Altogether, in the rat retina, the fate of intravenously injected ICG, which binds to lipoproteins and to albumin, reflects well the known kinetics of albumin and lipoprotein transports into the retina. 

The reason for ICG accumulation in mast cells can be understood by the fact that mast cells metabolize lipoproteins and rapidly internalizes LDL,³² particularly when activated.³³ Because not all mast cells were tryptase positive and not all tryptase-positive cells were ICG labeled, it cannot be excluded that several types of mast cells with various activation states are present in the choroid. Mast cells, enriched along large vessels in the choroid, play an important role in physiopathology and can induce subretinal fluid accumulation and inflammation upon degranulation.³⁴ ICGA should be reanalyzed in cases of choroidal inflammation with a specific focus on hyperfluorescent dots with the idea that such dots (around 50 µm) could represent activated mast cells. 

Finally, we have identified that ICG accumulates in choroidal nerves, particularly the large bundles, which reflects well the known entry of albumin into the endoneurial space through the endoneurial vasculature at 1.2 mg.g^–1.day^–1 and with a daily turnover of endoneurial albumin of approximately 30%.³⁵ Mata et al.³⁶ showed that serum albumin is found within the perineurium and endoneurium but not in the axon, suggesting that axons are exposed to serum proteins in normal nerves. The late phase ICGA hyperfluorescent dots and tracks found in the retinal periphery of eyes with transthyretin amyloidosis angiopathy²² might correspond, at least in part, with the large choroidal nerves running in the walls of choroidal vessels, mostly arteries, and not with vascular amyloid plaques because plaques in retinal vessels, clearly observed on fundus photography, are not stained by ICG at all the angiographic times of the sequence (Fig. 8). If such amyloid deposits were stained by ICG, retinal plaques should also be stained, because intramural deposits alter the endothelial integrity, giving access to ICG.³⁷ In addition, amyloid neuropathy, particularly alteration of nerves from the autonomous nervous system is part of the clinical signs of the disease³⁸ and increase of albumin entry in the endoneurium through endothelial alterations is an indicator of pathologic nerves.^37–39 ICGA is not commonly performed in patients with diabetic retinopathy, although it has been used to identify the early occurrence of choroidal pathology.⁴⁰ because diabetic patients with retinopathy often suffer from peripheral neuropathy, including autonomic neuropathy,⁴¹ late phase ICG might be useful to reveal choroidal neuropathy in diabetic eyes. 

The interpretation of clinical ICGA images from observations made on albinos rats is far from being straightforward, partly because these animals are albino and partly because the observation times are longer than those usually used in clinical examinations. However, the numerous hyperfluorescent or hypofluorescent ICG images that we observe but whose significance we do not fully understand should be analyzed with a new perspective, considering that ICG is a unique tool to follow the metabolism of proteins and lipoproteins in the eye and not only an inert dye. ICGA image interpretation should also consider that the fluorescence properties of ICG depend on the concentration of ICG when bound to proteins and that high local concentrations can induce hypofluorescence by quenching mechanisms. 

The observation we report herein open new perspectives in the use of ICG as a metabolic marker, but also a marker of the neural and immune components of the choroid. If ICGA could improve the identification of choroidal neuropathy in several retinal disease, it could help our understanding of disease and their treatments. Similar studies should be replicated in nonhuman primates for better translation to the clinic. 

Acknowledgments

The authors thank Theano Irinopoulou at the imaging platform of Fer à Moulin Institute (Paris, France) for its technical support. The authors also thank the Abraham J. & Phyllis Katz Foundation for their financial support for conducting research. 

Funded by the Agence Nationale de la Recherche (ANR-20-CE17-0034). 

Institutional Review Board Statement: The animal study protocol was performed in accordance with the European Communities Council Directive 86/609/EEC and French national regulations and approved by local ethical committees (#25158, Charles Darwin). 

Data Availability Statement: The data from the study are available from the corresponding author upon reasonable request. 

Disclosure: D. Mejlachowicz, None; P. Lassiaz, None; M. Zola, None; B. Leclercq, None; E. Gélizé, None; S. Achiedo, None; M. Zhao, None; A. Rousseau, None; F. Behar-Cohen, None