Toxicity Assessment of Intravitreal Triamcinolone and Bevacizumab in a Retinal Explant Mouse Model Using Two-Photon Microscopy

Abstract

Purpose.: Intravitreal drug administration leads to high intraocular concentrations with potentially toxic effects on ocular tissues. This study was an assessment of the toxicity of triamcinolone and bevacizumab in living retinal explants using two-photon (2P) microscopy.

Methods.: Wild-type mice received intravitreal injections of triamcinolone, bevacizumab, or vehicle. Ten and 45 days after injection, wholemounted retinal explants were incubated with the fluorescent dye sulforhodamine 101 (SR101) to analyze morphology and tissue damage with 2P microscopy ex vivo. Retinas that received the same treatment were stained for apoptosis (TUNEL) and glial activation (GFAP). An intravitreal injection of NMDA (N-methyl-d-aspartate) was used as a positive control to ensure the fidelity of detection of retinal damage with ex vivo 2P microscopy.

Results.: Overall retinal morphology was undisturbed after all procedures and time points. NMDA injection resulted in a strong increase in the number of SR101-labeled cells and increased apoptosis and glial activation when compared with sham-injected eyes. This result was in contrast to exposure to bevacizumab, which caused no appreciable damage. After triamcinolone treatment, marked damage in the inner retina was observed. However, damaged cells were restricted to sharply demarcated areas, and only mild changes in TUNEL-positive cells and GFAP activation was observed when compared to sham-injected eyes.

Conclusions.: 2P microscopy in combination with SR101 staining allows fast morphologic assessment of living retinal explants and can be used to evaluate adverse effects on retinal viability of test substances. Bevacizumab treatment did not cause any detectable retinal damage, whereas triamcinolone was associated with substantial, although spatially restricted, damage.

Since the beginning of this century, a paradigm change has taken place. The intravitreal cavity has no longer been considered to be untouchable, but it has been used in exponentially increasing numbers as a drug reservoir for the treatment of intraocular diseases, mainly exudative age-related macular degeneration and diabetic retinopathy. ¹ By using the intravitreal approach, it is feasible to achieve drug concentrations far exceeding the ones observed to date at the site of the required action, almost without systemic side effects. ¹  

Triamcinolone acetonide and bevacizumab are among the most commonly used drugs for intravitreal delivery. ^2,3 Triamcinolone is a steroid and, accordingly, possesses anti-inflammatory and anti-edematous properties. It has been the most widely injected drug to date for various intraocular neovascular and edematous diseases (reviewed in Ref. 1). Bevacizumab is a full-sized antibody directed against all isoforms of vascular endothelial growth factor (VEGF). Recent research has shown that VEGF plays an important role in many ocular diseases that involve neovascularization. Bevacizumab is one of several new anti-VEGF agents that have revolutionized the treatment of neovascular age-related macular degeneration and hold great potential for other blinding conditions such as diabetic retinopathy, retinopathy of prematurity, and neovascular glaucoma (reviewed in Ref. 4). Because these two drugs have been introduced into ophthalmic clinical practice almost without having performed large-scale, randomized, double-masked, controlled studies and hence are used as off-label therapy, information of their intraocular safety is often based on clinical case-series studies, and only recently has the evaluation of their properties been designed for a scientific evaluation of a relatively low scientific level. ⁵  

The purpose of our study was therefore to examine the potential toxicity of intravitreal triamcinolone and bevacizumab. Conceivable adverse effects include damage to or cell death of specific or all cell types, as well as inflammation and gross morphologic changes due to edema or disturbed vascularization. For the ex vivo assessment of retinal tissue, we used two-photon (2P) microscopy. ⁶ We made use of the fact that 2P microscopy, in combination with extracellular fluorescent dyes, such as sulforhodamine 101 (SR101), allows assessment of morphology as well as viability of the living retinal explant at subcellular resolution, ^7,8 which in principle can be consecutively used for functional measurements using optical and/or electrophysiological recording techniques. In this study, we obtained from living tissue 2P morphology data, which was complemented with conventional light microscopy-based paraffin histology and immunohistochemistry with markers for apoptosis and microglial activation. 

Methods

Animals, Intravitreal Injections, and Tissue Preparation

Adult wild-type mice (C57 black) aged 2 to 3 months were used. The study adhered to the statement of the ARVO for the Use of Animals in Ophthalmic and Vision Research. In addition, all experimental procedures were approved by the appropriate state authority (Regierungspräsidium Karlsruhe, Germany). 

Mice received an unilateral intravitreal injection of either triamcinolone, bevacizumab or NMDA (N-methyl-d-aspartate) or a sham injection of vehicle only (physiological NaCl solution). The injections were performed in a standardized manner in anesthetized mice. After penetration of the sclera in a slightly oblique direction, the beveled needle tip of a microsyringe (32-gauge; Hamilton, Bonaduz, Switzerland) was placed intravitreally in the retrolental space of the eye under direct ophthalmoscopic control with an operating microscope. During the placement of the needle, some vitreous outflow occurs due to the beveled needle configuration, which greatly reduces the risk of subsequent drug reflux. In our hands, 2 to 3 μL can be reliably injected intravitreally without reflux during the injection or apparent postoperative inflammation in the mouse eye. Therefore, we injected one eye with up to 3 μL of the test substance, whereas the contralateral eye was injected with up to 3 μL physiological NaCl-solution (sham injection) or was left uninjected. Note that because of the preinjection vitreous outflow, the injected volume did not add to but replaced fluid in the eye and, therefore, for calculating the final (maximum) drug concentrations, the standard vitreous volume of the mouse eye (7 μL) was used. 

Triamcinolone acetonide (Volon A; Bristol-Myers-Squibb, München, Germany) was prepared by filtration of an ampoule containing 40 mg triamcinolone in 1 mL benzene alcohol, thus removing the supernatant and replacing it with Ringer's solution as described in. ⁹ This procedure was repeated twice. Usually the injected amount is 25 mg triamcinolone in 0.3 mL corresponding to a concentration of 83.333 mg/mL, thus 3 μL of the solution contain 250 μg of triamcinolone. With a vitreous cavity of 7 μL in mice, ^10,11 a theoretical triamcinolone concentration of 35.714 mg/mL was achieved. This is ∼5.7-times the concentration of triamcinolone when compared to the high-dosage intravitreal triamcinolone therapy of 25 mg per injection in patients (25 mg triamcinolone per 4-mL vitreous cavity or 6.25 mg/mL). ⁹ With this protocol, we aimed at obtaining the highest possible dose for this toxicity study; however, the actually achieved concentration in the vitreous of the treated mice is likely to be somewhat lower due to variability in drug preparation and injection procedure. 

Similarly, bevacizumab (Avastin; Roche, Grenzach-Wyhlen, Switzerland) was obtained in a concentration of 25 mg/mL. An injected volume of 3 μL equals a dosage of 75 μg bevacizumab, resulting in an intravitreal concentration of 75 μg per 7 μL vitreous volume in mice, or 10.714 mg/mL. This concentration is theoretically ∼28.6-times higher than the concentration of bevacizumab when intravitreally injected in humans at a standard dosage (1.5 mg bevacizumab in 4 mL vitreous cavity or 0.375 mg/mL). 

For N-methyl-d-aspartate (NMDA; Sigma-Aldrich, München, Germany) we injected 2 μL of a 4-nM solution, which was shown to be sufficient to induce ganglion cell apoptosis. ¹²  

At 10 days and 45 days after the intravitreal injections of bevacizumab or TA and at 1 to 2 days after the injection of NMDA, the mice were killed by cerebral dislocation. The eyes were quickly enucleated in Ames' medium (Sigma-Aldrich) for ex vivo 2P microscopy, or they were fixed in 4% paraformaldehyde for paraffin histology. All procedures were performed at room temperature and at light-adapted conditions. 

For the ex vivo examination of the retina by 2P microscopy, the ocular anterior segment including the lens were removed. The retinas were then separated from sclera and retinal pigment epithelium and kept in oxygenized (5% CO₂, 95% O₂) Ames' medium. 

2P Microscopy

Two photon microscopy ⁶ has been described elsewhere in detail, ^13,14 including its application in the retina. ^7,8 Therefore, we focus here on aspects relevant to the presented experiments. We examined the retinal wholemounts with a custom-built, upright, 2P microscope ⁸ equipped with a water-immersion lens (20 × 0.95NA XLUMPlanFI; Olympus, Tokyo, Japan, or 16 × 0.8NA LWD DIC N2; Nikon, Tokyo, Japan) and two detector channels, of which one channel (emission filter HQ 622/36m-2p-18deg; Chroma/AHF, Tübingen, Germany) was used to detect fluorescence. The 2P excitation source was a mode-locked Ti/Sapphire laser (Mira-900; Coherent, Dieburg, Germany) tuned to a ∼930-nm wavelength. 

For the examination, four radial relaxing incisions were made in the retina to enable flatmounting on a filter paper (0.8 μm black, AABP; Millipore, Schwalbach/Ts., Germany). ¹⁵ All preparations were performed under immersion with oxygenated Ames' medium. During handling of the retinal tissue, care was taken to do as little mechanical damage as possible, although typical artifacts were unavoidable (see Fig. 2B). The filter paper with the retina was placed in the recording chamber on the stage of the 2P microscope, where the tissue was continuously perfused with oxygenated Ames' medium. 

The interstitial space was labeled by adding low concentrations of sulforhodamine 101 (SR101, 3–7 μM; Sigma-Aldrich) to the extracellular medium, providing a very detailed image of the intact retina with subcellular optical resolution. ^7,8 Counterstaining with SR101, which is nontoxic, ¹⁶ allows identification of cell classes and even some cell types based on morphologic properties, including soma size, shape, and position within the retinal layers. ⁸ SR101 labeling also allows assessment of the viability of the tissue, because as a hydrophilic dye, it does not penetrate intact membranes of living cells. ⁸ Some types of retinal neurons, including photoreceptors, bipolar cells, and possibly some types of amacrine cells, take up SR101 into their processes but usually not into their soma. ⁸ For example, SR101 labels the synaptic terminals of rod bipolar cells and photoreceptors, but not their somata. Such staining likely reflects active uptake due to vesicle recycling. ¹⁶ When cells become leaky due to damage, SR101 enters and renders these cells (including their somata) brightly fluorescent (reviewed in Ref. 8). 

2P image stacks of the SR101-stained retinas were obtained in at least four areas of each retina equidistant to the optic disc. Retinal regions with mechanical damage (e.g., close to the radial incisions or imprints of forceps in the far periphery) caused by the tissue preparation were avoided. Overview stacks (512 × 512 pixels, with 237- to 564-μm side length) as well as close-ups (256 × 256 pixels, with 59- to 141-μm side length) were recorded (in 0.5 to 2.0 μm z-steps, two to three averages per frame) using the microscope's acquisition software (CfNT, developed by Ray Stepnoski, Bell Laboratories, Murray Hill, NJ, and Michael Müller, Max Planck Institute [MPI] for Medical Research, Heidelberg, Germany). Off-line processing of the image data was performed with ImageJ (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html, with the CfNT import plug-in and reader available in the public domain at http://www.mpimf-heidelberg.mpg.de/abteilungen/biomedizinischeOptik/software/cfdPluginImageJ/index.html/, MPI for Medical Research). 

Histologic Processing and Conventional Microscopy

A separate batch of eyes injected with bevacizumab, triamcinolone, or NMDA were processed for standard histology. Serial sagittal histologic sections of a thickness of 12 μm running through the pupil and the optic nerve head were obtained. One set of sections was stained for apoptosis (TUNEL) by using a standard kit (TACS-XL-Basic, In Situ Apoptosis Detection Kit; R&D Systems, Minneapolis, MN) and 3,3′-diaminobenzidine (DAB) labeling. To avoid masking, counterstaining was not performed. Lymph nodes tissue and nuclease-treated retinal tissue slides served as the positive control. Another set of sections was stained for glial fiber acidic protein (GFAP) in a standard protocol with a rabbit polyclonal anti-GFAP-antibody (Novus Biologicals Inc., Littleton, CO) at a dilution of 1:500. The primary antibody was visualized with 3-amino-9-ethylcarbazole (AEC), resulting in a brownish red stain. Again, to avoid masking, no counterstaining was performed. Histologic sections of mouse brain served as the positive control (not shown). 

All slides were analyzed using a light microscope (Diaplan Details, Leica, Wetzlar, Germany) with digital image capturing (Digital Camera DC 500; Leica). 

Data Analysis

To determine the cell density in the ganglion cell layer (GCL), we counted the somata in 1 to 4 fields (between 74 × 78 and 450 × 474 μm²) per animal. We selected those frames of the 2P microscopic image stacks with the focal plane in the center of the GCL. Fields with SR101 cell labeling too intense to allow resolution of unlabeled cells were excluded. The GCL cell densities estimated from our counts (Table 1) matched well the densities previously reported for the mouse retina. ¹⁷ For this comparison, we plotted the published GCL cell densities as a function of retinal eccentricity and fitted a Gauss curve,  

 

with density D (in square millimeters), eccentricity x (distance from the optic disc in micrometers), and the coefficients a = 10,017, b = −4,002.3, c = 1,916.8, and w = 713.8. The respective fit parameters for the inner nuclear layer (INL) cell density were a = 80,568, b = 30,338, c = 723.4, w = 1,005.4. We then compared our ganglion cell layer cell density measured for a certain eccentricity with the density predicted by the Gauss fit (see equation) for the same eccentricity and found no statistically significant difference (Wilcoxon signed test, n = 17 pairs: P ≥ 0.32). In the light of this good agreement, we used average densities determined with the equation (for the eccentricity range from 415–1533 μm) to estimate the percentage of damaged cells in GCL and INL (Table 2). Note that cell density measurements included all somata in the respective layers: ganglion and displaced amacrine cells in the GCL, and amacrine, bipolar, horizontal, and Müller cells in the INL. 

Table 1.

 

Total Density of Cells in the GCL

Condition	Total Density (GCL), mm−2 (Fields, n)	P vs. Triamcinolone	P vs. Bevacizumab
Control NaCl	8801 ± 1579	0.11	0.07
	(14)
Triamcinolone	7952 ± 855		0.76
	(9)
Bevacizumab	7506 ± 1619
	(10)

 For the number of animals and eyes, see Table 2. P determined by t-test.

Table 2.

 

Density of SR101-Filled Cells in the GCL and INL

Condition	Eyes (Mice)	GCL			INL
		Density, mm−2 (Fields, n)	% of All Cells (Eyes, n)	P vs. Control	Density, mm−2 (Fields, n)	% of All Cells (Eyes, n)	P vs. Control
Control NaCl	6	100 ± 231	1.1 ± 1.7		72 ± 307	0.1 ± 0.2
	(6)	(29)	(6)		(29)	(6)
Triamcinolone total	9	1647 ± 2605	13.7 ± 14.6	<0.01**	9424 ± 22881	10.7 ± 10.5	0.02*
	(6)	(43)	(9)		(35)	(9)
2 wk	5	741 ± 2013	8.2 ± 10.8	0.15	7530 ± 22102	7.4 ± 10.4	0.14
	(4)	(22)	(5)		(21)	(5)
4 wk	4	2596 ± 2854	20.6 ± 17.2	<0.01**	12266 ± 24561	14.7 ± 10.5	0.09
	(2)	(21)	(4)		(14)	(4)
Bevacizumab total	6	98 ± 259	1.3 ± 2.0	0.97	95 ± 207	0.1 ± 0.2	0.75
	(5)	(25)	(6)		(22)	(6)
2 wk	4	139 ± 307	1.8 ± 2.4	0.65	149 ± 245	0.1 ± 0.2	0.38
	(3)	(17)	(4)		(14)	(4)
4 wk	2	9 ± 18	[0.2, 0.1]	0.04*	0 ± 0	[0.0, 0.0]	0.22
	(2)	(8)	(2)		(8)	(2)
NMDA (24–48 h)	4	1493 ± 773	15.7 ± 5.9	<0.01**	815 ± 1228	0.8 ± 0.4	<0.01**
	(3)	(26)	(4)		(23)	(4)

 SR101-filled cell fractions (in %) calculated per eye by using average cell densities estimated with the equation in the text for the eccentricity range 415–1533 μm (GCL, 9,008 mm⁻²; INL, 106,658 mm⁻²). For n <3 fields, individual data instead of averages are given in parentheses; P determined by t-test (* P < 0.05; ** P < 0.01).

To evaluate the anatomic structure and tissue integrity from the 2P image stacks, we determined the density of SR101-filled somata in GCL and INL (Table 2) as well as the thickness of the different layers (described later; Table 3). SR-filled cells were counted in at least 8 fields (two eyes; up to 43 fields/nine eyes for the GCL and 35 fields/nine eyes for the INL) per injected drug and time point. For ease of reproducibility and recognition, GCL and inner plexiform layer (IPL) thicknesses were not separately measured but in combination. The INL and outer nuclear layer (ONL) were measured separately, and the vessel plexus was omitted. The number of frames per layer was counted and converted to micrometers corresponding to the layer thickness. 

Table 3.

 

Retinal Layer Thickness for Control Condition and after Injection

Condition	Eyes (Mice), n	Fields (n)	GCL + IPL (P vs. Control)	INL (P vs. Control)	ONL (P vs. Control)
Control	6	12	26.3 ± 9.8	23.4 ± 6.3	57.7 ± 12.6
	(6)
Triamcinolone	5	9	23.3 ± 4.2	18.9 ± 2.7	54.4 ± 9.9
	(5)		(0.36)	(0.04*)	(0.52)
Bevacizumab	10	52	23.2 ± 6.0	20.8 ± 5.8	53.8 ± 10.4
	(10)		(0.31)	(0.21)	(0.34)
NMDA	3	25	19.9 ± 5.1	17.8 ± 3.9	48.4 ± 13.8
	(2)		(0.05)	(0.01**)	(0.06)

  P determined by t-test (* P < 0.05; ** P < 0.01).

Results

2P Microscopy

We investigated possible side effects of intravitreal injection of triamcinolone and bevacizumab. To differentiate between drug-specific effects and procedural damage we tested sham-injected and untreated eyes. Ten and 45 days after the saline injection, retinal explants were stained with SR101 and the tissue morphology was recorded with 2P microscopy (Fig. 1). The morphology of sham-injected retinas was indistinguishable from retinas dissected from untreated eyes (not shown). However, in both cases, we did observe typical dissection-related damage, such as an increased density of SR101-stained cells around the optic nerve head (Fig. 2A) or where the tissue was mechanically touched, which happened only at the margin of the retina (Fig. 2B). In most areas, the tissue was healthy, displaying only a few isolated SR101-positive cells mostly in the GCL and the INL (Fig. 2C, arrowheads). This suggests that the intravitreal injection itself does not cause any detectable damage. 

Figure 1.

 

2P micrographs of living mouse retina (sham injection, intravitreal application of NaCl solution) stained with SR101. In the intact retina, the dye stains the extracellular space such that cells appear as dark profiles. In addition, blood vessels and the synaptic endings of a few cell types (most prominently photoreceptor terminals, see A/OPL) are stained by SR101. (A) Low-magnification overview showing the different retinal layers or borders between layers. Images show the same field of view at different depths (from a stack with 2-μm steps between frames; each image is the average of two consecutive frames). Examples of erythrocytes (arrowheads) within blood vessels and nerve fiber bundles (*) are indicated in the first image of the series. No damaged cell somata were visible. In addition to the interstitial space, SR101 labeled structures including blood vessels, photoreceptor terminals (see OPL frame), and some photoreceptor outer segments (see OS frame). (B) Higher magnification view (different retina than in A; from a stack with 0.5-μm steps between frames; each image is the average of two to three consecutive frames). Scale bar, 20 μm.

Figure 2.

 

2P micrographs showing typical tissue damage observed in sham-injected retina. Bright fluorescent labeling of cells with SR101 is indicative of damaged tissue. (A) SR101-labeled cells were observed in an area with a diameter of ∼200 to 400 μm around the optic disc, probably due to mechanical stress when the optic nerve was cut during the dissection. (B) Example of locally restricted damage caused by touching the tissue (e.g., with the tips of a forceps during the dissection). (C) Examples for SR101-labeled (arrowheads), in an otherwise normal and healthy retina. A few ganglion cell somata (GCL) with their dendrites (IPL) and a single bipolar cell (INL/OPL) were SR-stained (the dim, regularly distributed structures are cone pedicles, cf. Fig. 1B at 79 μm). Scale bar, 50 μm.

After the intravitreal injection of triamcinolone, 2P micrographs showed a normal overall retina structure and layering, but revealed a high number of SR101-positive cells in both the GCL and INL (Fig 3A), in sharply demarcated areas (Fig 3C), whereas other quadrants of the same eyes appeared largely healthy (Fig 3B). After injection with bevacizumab, 2P microscopy did not reveal any discernable differences compared with sham-injected retinas (Fig 3D). As a positive control, we intravitreally injected mice with NMDA, which is known to induce apoptosis in retinal ganglion cells. ¹² In retinal explants dissected at 24 to 48 hours after the intravitreal injection, we found a high proportion of SR101-positive cells (mostly in the GCL, to a lesser extent in the INL) uniformly distributed in all quadrants (Figs. 3E, 3F). 

Figure 3.

 

2P micrographs showing retinal tissue after ocular injection with triamcinolone (240 μg), bevacizumab (75 μg), and NMDA (8 nM). (A–C) Four (A, B) or 2 (C) weeks after triamcinolone (injection large retinal areas contained a high percentage of SR101-filled cells in both the GCL and the INL, suggesting substantial tissue damage. Other areas were almost free of SR101-labeled cells. (B arrowhead) SR101-filled ganglion cell). Sharp borders were observed between SR101-labeled and apparently normal areas (C). (D) Retinal tissue from bevacizumab-injected eyes were not significantly different from control tissue (cf. Fig. 1). (E–F) Many brightly SR101-labeled in both GCL and INL were observed 24 to 48 hours after intravitreal injection of NMDA. The labeling varied among regions of the same retina (cf. E, F). Scale bar, 50 μm.

Damaged cells are visible during evaluation, whereas possible cell loss that occurred before evaluation was not accounted for by SR101 staining. Therefore, we measured the total cell density in the GCL, the density/fraction of SR101 positive cells in the GCL and the INL, and the thickness of the various retinal layers. 

The total density of cells in the GCL is an indicator for retinal damage, resulting from cell loss and altered tissue integrity and microenvironment. To examine whether the treatment with triamcinolone or bevacizumab resulted in cell loss in the GCL, we determined the total GCL cell density. Since the incubation time was considerably shorter in the eyes injected with NMDA, these eyes were not included in this analysis. Although the total cell density in the GCL tended to be lower in the triamcinolone-injected eyes and the bevacizumab-treated eyes than in the control eyes, these differences were not statistically significant (Table 1). 

Next, we determined the density of SR101-positive, and thus presumably damaged, cells in the GCL and the INL for the different drugs injected and for the different incubation times. In the control retinas, the fraction of SR101-positive cells was on average ∼1% in the GCL and ∼0.1% in the INL (for all densities and statistics see Table 2, Fig. 4A). For the triamcinolone-treated eyes, the fraction of SR101-positive cells was markedly increased in both the GCL (∼14%) and the INL (∼11%). These differences were statistically significant when the data from the two time points were combined. However, due to the large variations within the group of eyes treated with triamcinolone (Fig. 3C) and within different fields of the same retina (Fig. 4A), the increase in SR101-positive cells was only significant in the GCL when the triamcinolone treatment lasted 4 weeks (Table 2). The bevacizumab-treated eyes did not vary significantly from the control eyes in the average fraction of SR101-positive cells, for both the GCL and the INL. Of interest, the average density of SR101-positive cells in the GCL showed a small but not statistically significant decrease after 45 days of bevacizumab treatment (Table 2). Intravitreal NMDA-injection was, as expected, associated with a significant increase in the percentage of SR101-positive cells (GCL: ∼16%, INL: ∼0.8%). 

Figure 4.

 

Densities of SR101-filled cells and thickness of retinal layers. (A) Density of SR101-labeled cells in the GCL (top) and INL (bottom) across the examined eyes (n = 25). For both cell layers, the same histogram is plotted twice but with different scales, to allow appreciation of the density difference between conditions. Large error bars result from the inhomogeneous distribution of SR101-labeled cells in some retinas (cf., for example, Figs. 3A–C). (B) Thickness of retinal layers (ONL, INL, and GCL+IPL) was averaged across the different conditions (for number of eyes and animals see Table 3).

We also determined the thickness of the different retinal layers, which can be an indicator for intraretinal edema or tissue disintegration. From z-stacks, we measured the thickness of the ONL and INL and, combined in one measurement, the IPL and GCL in the control eyes and the eyes treated with the different drugs. The retinas from the eyes treated with triamcinolone, bevacizumab, or NMDA tended to be thinner, however, except for a small decrease in INL thickness for the triamcinolone group and the NMDA-injected eyes, we found no statistically significant differences (Fig. 4B, Table 3). 

TUNEL and GFAP Histology

To back up the data from 2P microscopy, we performed conventional light histology on control eyes as well on the eyes treated with triamcinolone, bevacizumab, and NMDA after the different incubation times, and we tested for microglia activation via GFAP immunohistochemistry (Figs. 5A–D) and apoptosis via TUNEL (Figs. 5E–I). 

Figure 5.

 

Paraffin histology. (A–D) Vertical sections stained using an antibody against GFAP. Retinas from eyes injected with NaCl (A), bevacizumab (B), triamcinolone acetonide (TA; C1, C2), and NMDA (D). (E–I) Vertical sections labeled for apoptosis (TUNEL); arrowheads: apoptotic cells. Control section treated with nuclease (E). The other sections show eyes injected with NaCl (F), bevacizumab (G), TA (H), and NMDA (I1 and I2). Incubation times: NaCl, bevacizumab, and TA: 4 weeks: NMDA 2 days. Scale bars, 25 μm.

Compared to the control eyes (Fig. 5A), we did not detect an increase in GFAP-labeling in the bevacizumab-treated retinas. In contrast, moderate GFAP-labeling was detected in some samples of triamcinolone-treated eyes (Fig. 5C1) but not in others (Fig. 5C2), suggesting regional but appreciable microglial activation. Both results are consistent with those of the SR101 labeling, which predicted substantial but “patchy” tissue damage with triamcinolone but none with bevacizumab. Also consistent with the SR101 data were the marked staining for GFAP in the NMDA-treated tissue (Fig. 5D). 

Only a few TUNEL-positive cells (0–1 cells per field, usually in the ONL) were found in the control retinas (Fig. 5F; compare with nuclease-treated tissue in Fig. 5E) and in retina treated with bevacizumab (Fig. 5G). For the triamcinolone group, moderate differences in the TUNEL-positive cell count (1–4 cells per field, usually in the ONL, Fig. 5H) compared with the control eyes were detected. Also in the NMDA-injected eyes, the count of TUNEL-positive cells was increased (4–6 per field, mostly in the GCL and the INL). Because of the low number of positive cells, a statistical analysis could not be performed. 

Discussion

In this study, we applied 2P microscopy ⁶ in combination with SR101-labeling of cells ^7,8 to evaluate retinal toxicity of intravitreally injected drugs. We used the fact that in healthy tissue the hydrophilic SR101 stains the interstitial space and diffuses into intact cells. However, when cell membranes become compromised, SR101 enters the cells, which become brightly labeled (see the Methods section). In addition to the possibility of a rapid examination of the tissue, an important advantage of 2P microscopy compared with confocal or conventional fluorescence microcopy lies in the opportunity to study the light-sensitive retinal explant ex vivo, made possible by 2P microscopy's use of an infrared laser as the excitation source, thus, avoiding photopigment bleaching. ⁷ Therefore, in addition to analyzing the retinal explant morphologically (as described in this study), 2P microscopy can be further used for physiological measurements of light-evoked activity in electrophysiological processes or optical recording techniques such as Ca²⁺ imaging. ^7,18–21 Furthermore, with 2P microscopy, fluorophore excitation is restricted to a minuscule focal volume ^6,14 and, thus, phototoxic effects are usually less pronounced than with microscopic techniques relying on single-photon excitation. In line with this, an increase in the number of SR101-labeled cells during 2P scanning was not detected. 

For the present pilot study, we used two of the currently most frequently intravitreally administered drugs in clinical ophthalmology: triamcinolone and bevacizumab. As a positive control, we intravitreally injected NMDA, a glutamate receptor agonist, which induces ganglion cell apoptosis and causes apoptotic changes in the inner retina. ¹² After the injection of NMDA, we detected a uniform distribution of SR101-positive, and thus presumably damaged cells mainly in the GCL. The light microscopic histology, as the conventional method for the evaluation of tissue damage, confirmed the induction of apoptosis and glial activation after intravitreal treatment with NMDA. To our knowledge, we showed in this work for the first time that 2P microscopy in combination with fluorescent labeling allows assessment of retinal damage in response to drug treatment in the living retinal explant. 

Both drugs used, triamcinolone and bevacizumab, are clinically used off-label, and for both, the issue of biocompatibility and toxicity to the different intraocular tissues has been raised and studied with partially disparate results. ^22,23 To date, there have been no in vivo studies that emulate the clinical administration of triamcinolone or bevacizumab with longitudinal evaluation of adverse effects. 

One of the major concerns with triamcinolone has been the solvent agent benzyl alcohol in the commercially available solutions, since several reports have recommended removal of the solvent agent. ^24,25 When applied to cultured primary rat retinal cells, triamcinolone induces oxidative injury in a glucocorticoid receptor-independent manner, suggesting toxic potential. ²⁶ Furthermore, epiretinal triamcinolone crystals have long been suspected of causing retinal damage. In a model of perfused bovine retina, massive ganglion cell damage has been reported, whereas electroretinographic amplitudes remained unchanged. ²⁷ In direct contact with retinal cells in cell culture, triamcinolone crystals have been observed to cause cell damage, ²⁸ perhaps due to the lack of the potentially protective effect of the ILM or vitreous. This observation may explain why up to now, a toxic effect of intravitreally applied triamcinolone with the solvent agent removed has not been reported, despite its widespread use. ²⁸ In the present study, eyes treated intravitreally with triamcinolone in high dosages (up to 5.7 times higher than the clinical dose) showed sharply demarcated areas with retinal cell damage, with other parts of the retina appearing to be completely unchanged compared with control eyes. One possible explanation for this is that the triamcinolone crystals lying on the retinal surface were associated with the strictly localized retinal changes, whereas solved triamcinolone did not cause diffuse change in the retina. In agreement with this, in triamcinolone-treated eyes, increased glial activation was observed in some but not all eyes. Why an accompanying slight increase in apoptosis was not found is unclear. 

Because bevacizumab is an anti-VEGF agent, the major concern has been an alteration of the physiological vasculature, since VEGF has been reported to play an integral part in retinal vessel physiology. Previous experimental investigations did not find changes in cell viability or apoptosis rates in cultured retinal cells in avascular cell cultures with bevacizumab (e.g., Ref. 29). Correspondingly, in the present study, we did not observe any significant abnormalities at any time point in eyes that received bevacizumab intravitreally in high dosages (up to 28.6 times higher than the clinical dose). For both drugs, high concentration levels were chosen to ensure that subclinical changes may become appreciable. With no alterations seen at high concentrations, it is conceivable that bevacizumab is also safe at lower concentrations. 

Systematic analysis of retinal layer thickness did not show grave changes except for a small decrease in INL thickness for the triamcinolone group and the NMDA-injected eyes. Given that healthy wild-type animals with no primary retinal changes were included in the study, a treatment effect like a reduction in retinal edema, as seen in patients, was not to be expected. However, the lack of a toxic effect could be demonstrated. Such an effect could have resulted in either an inflammatory response with tissue activation including cellular and fluid influx with resulting retinal thickening or alternatively cell death and consequently marked thinning of the nuclear layers. 

There are certain limitations to our study relating to the life stain with SR101. This stain does not allow differentiation between irreversible and transient damage. It is conceivable, that SR101 enters cells after transient damage but is then extruded when the cells recover. Some activity-dependent uptake of SR101 into vesicles has been reported for motor neuron endings, ^30,31 but not for neurons of the central nervous system. ¹⁶ In the neocortex, SR101 is specifically taken up by astroglia. ³² In the retina, some types of retinal neurons, including photoreceptors, bipolar cells, and possibly also some amacrine cells, take up SR101 even when they are healthy. ⁸ However, in these cell groups, SR101 typically accumulates in synaptic terminals but not in their somata, and therefore it is unlikely that these are confused with cells stained with SR101 due to damage (see the Methods section). 

In conclusion, intravitreal bevacizumab did not lead to any severe morphologic changes in the retina, as observed in the retinal explant ex vivo with 2P microscopy. Eyes with intravitreal triamcinolone showed, within the otherwise completely normal-appearing retina, patchy areas with cell damage, possibly due to triamcinolone crystals in contact with the retina in these locations. Furthermore, our study shows that 2P microscopy is a useful tool for ex vivo assessment of retinal explants. In particular in view of the upcoming plethora of novel drugs for intravitreal application, 2P imaging may be a valuable tool in assessing risk profiles. 

Footnotes

 Presented at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2007.

Footnotes

 Supported by the Max Planck Society.

Footnotes

 Disclosure: F.C. Schlichtenbrede, None; W. Mittmann, None; F. Rensch, None; F. vom Hagen, None; J.B. Jonas, None; T. Euler, None

Footnotes

 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.

The authors thank Elizabeth Batton for counting cells and Winfried Denk for his support.