In Vivo Imaging of Retinal Ganglion Cell Axons within the Nerve Fiber Layer

Abstract

Purpose.: Optic nerve injury causes loss of retinal ganglion cells (RGCs) and their axons. The reduction in RGC counts over time in axonal injury is well studied, but the correlation with the timing of anterograde and retrograde axonal degeneration is less clear. The authors longitudinally imaged RGC axons stained with a chloromethyl derivative of fluorescein diacetate (CMFDA) in live rats after optic nerve injury.

Methods.: Optic nerves were transected. Three days later CMFDA was intravitreously injected. Confocal scanning laser ophthalmoscopy was performed daily, and mean fluorescence intensity and the number of CMFDA bundles were calculated. RGC soma survival was studied after retrograde fluorescence labeling. Retinal nerve fiber layer (RNFL) thickness was evaluated histologically.

Results.: CMFDA-positive RGC axon bundles could be imaged in vivo. Axons lost 68% ± 29% of their fluorescence by 7 days after transection compared with 25% ± 21% in nontransected eyes. The number of labeled axon bundles decreased by 61% ± 28% at 7 days after transection compared with 26% ± 9% in nontransected eyes. The number of retrograde-labeled RGCs detected in vivo declined by 53% at 7 days and by 76% at 14 days after transection. RGC soma and CMFDA axon counts decreased most rapidly between 5 and 7 days after transection. Histologic examination demonstrated a reduction in RNFL thickness 7 days after transection.

Conclusions.: Intravitreal CMFDA can be used to longitudinally monitor RGC axons within the RNFL in vivo. Imaging the disappearance of retrograde-labeled RGC somas and axons indicates that axonal and somal degeneration occur in parallel after axotomy.

Most optic neuropathies are axogenic diseases characterized by injury to RGC axons in the optic nerve, followed by progressive loss of RGC somas. In many cases the site of injury is obvious (e.g., compressive optic neuropathy or optic neuritis), whereas in more complex diseases such as glaucoma, ¹ histologic ² or psychophysical ^3,4 techniques demonstrate that axons are affected earlier than RGC somas. Assessment of retinal nerve fiber layer (RNFL) thickness is helpful in assessing the degeneration of RGCs in patients with optic neuropathies, such as optic neuritis, ⁵ compressive neuropathy, ⁶ and glaucoma. ⁷ RNFL thickness measurements using commercially available instruments make it possible to estimate the effect of RGC degeneration in clinical settings. 

Although the reduction in RGCs after axonal injury is well studied, the mechanisms and time course of anterograde and retrograde axonal degeneration are less well understood. The cellular and molecular mechanisms underlying axonal pathology have been primarily assayed by collecting fixed retinal or optic nerve samples in which axonal tracts have been marked by tracing techniques or immunohistochemistry. With the use of this classical approach, the number of axons that have been lost can then be quantitated, but only at single time points. ⁸ In vivo imaging provides a potential solution for such problems because it allows identified axons to be followed over time. Experimental models can be used to longitudinally quantitate RGC somal degeneration after optic nerve injury in rats ⁹ and mice. ^10,11 Similarly, the longitudinal reduction of RNFL thickness can be quantitated in vivo after optic nerve injury using confocal scanning laser ophthalmoscopy (CSLO) ¹² or optical coherence tomography (OCT). ¹³ However, these are bulk measures of RNFL thickness and do not distinguish individual bundles of RGC axons. 

In this study, we labeled individual RGC axonal bundles with a long-lasting and stable chloromethyl derivative of fluorescein diacetate (CMFDA) in the living rat and serially imaged them by CSLO after optic nerve transection. We observed that the time course of axonal degeneration paralleled that of RGC soma degeneration. 

Methods

Animals

All experiments were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and with the Canadian Council on Animal Care and were approved by the Animal Care Committee of the Maisonneuve-Rosemont Hospital Research Centre. Female Long-Evans rats, each weighing between 225 and 250 g, were purchased from Charles River (St-Constant, QC, Canada). 

Materials

Green CMFDA, 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide (DiR), and Alexa Fluor 488-dextran (10 kDa) were purchased from Invitrogen (Eugene, OR). Rhodamine-labeled Griffonia simplicifolia lectin I (GSL-I) was from Vector Laboratories (Burlingame, CA). 

Intrasheath Optic Nerve Transection

Rats were anesthetized with ketamine (50 mg/kg) and xylazine hydrochloride (100 mg/kg). All optic nerve transections were performed on the right eye using a Zeiss operating microscope. Limited lateral canthotomy was performed. The conjunctiva was then incised at the limbus, and the sub-Tenon space was bluntly dissected posteriorly. A small incision of the meningeal sheath was made 2 mm posterior to the globe. The optic nerve was then transected within the meninges with the bevel of a 23-gauge needle, under direct visualization. This procedure spares the meningeal vessels that carry the arterial circulation to the retina. Preservation of the retinal circulation was confirmed by indirect ophthalmoscopy. The skin was then closed with sutures, and erythromycin/polymyxin B/bacitracin antibiotic ointment was applied to the wound. 

Axonal Labeling with CMFDA

Intravitreal injections were performed immediately posterior to the pars plana using a 10-μL syringe attached to a 32-gauge needle (Hamilton Company, Reno, NV). CMFDA (4 μL) was slowly injected over 1 minute through the sclera at a 45° angle. This route of administration avoids retinal detachment or injury to the lens. Assuming the vitreous volume of an adult rat eye to be approximately 56 μL, ¹⁴ the final intravitreal concentration of CMFDA was approximately 60 μM. CMFDA was always injected superotemporally. Ophthalmic ointment containing erythromycin was applied to the globe after injection. 

Confocal Scanning Laser Imaging

CSLO retinal imaging was performed daily under ketamine/xylazine anesthesia. Pupils were dilated with one drop each of phenylephrine hydrochloride 2.5% and atropine sulfate 1%. CMFDA images were obtained using the 30° field of view and the automatic real-time mode on the HRA2 CSLO (Heidelberg Engineering, Heidelberg, Germany) at a sensitivity setting (93%) that maximized the signal/noise ratio. Four images per retina were obtained as shown in Figure 1. The first (F1) had the optic nerve head (ONH) at the center of the image. The second (F2) had the ONH at the edge of the image. The third (F3) was of the peripheral retina. The fourth (F4) was in between the F2 and F3 images. These four images were taken at the same retinal locations each session. 

Image Analysis

Axon bundle fluorescence intensities were calculated from the peripapillary F1 images, where the axons were closer together and the signal/noise ratio was high. Axon bundle counts were calculated from the peripheral F2 to F4 images, where the individual bundles could more easily be discerned. 

To subtract background, F1 was horizontally divided at the center of the ONH into two parts using ImageJ software (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html; Fig. 1A). The fluorescence intensity of CMFDA was calculated as (mean signal in the superior hemifield − mean signal in the inferior hemifield) × area of the superior hemifield. 

For axon bundle counting in images F2 to F4 (Figs. 1B, 1D, 1E), a line was drawn between two vessel bifurcations (as fiduciary points) that traversed CMFDA-stained axon bundles (Fig. 1D). The bundles on the line were manually counted in masked fashion. The line was drawn at the same place in each session. Fluorescence intensity and number of axon bundles were all normalized to the values measured the first day after CMFDA injection. 

Retrograde Labeling and Quantification of Retinal Ganglion Cells

RGCs were retrograde-labeled by stereotactic injection of a fluorescent tracer into the superior colliculus. Rats were placed in a stereotactic apparatus (Narishige Co. Ltd., Tokyo, Japan), and the skin of the skull was incised. The brain surface was exposed by drilling the parietal bone to facilitate dye injection. Fluorescent dyes were Alexa Fluor 488-dextran (20 mg/mL) and DiR (20 mg/mL). The dye solution (2.1 μL) was injected bilaterally 6.0 mm caudal to bregma and 1.5 mm lateral, to the midline to a depth of 4.5 mm from the skull surface, in accordance with preliminary experiments to identify the position of the superior colliculus. 

To quantitate the reduction in RGC counts after optic nerve transection, we imaged the retinas of eyes labeled by Alexa Fluor 488-dextran with the CSLO. Images of four retinal quadrants immediately adjacent to the ONH were obtained for each session. Because it was difficult to image the retina exactly the same way at each session, we counted RGCs within an area delineated by retinal vessels. Labeled cells in that area of each image were manually counted in a masked fashion by the same investigator. At least 100 RGCs were counted in each area. The relative number of labeled cells in the selected retinal area was calculated with respect to the number in nontransected nerves, with the density before optic nerve transection set at 100%. 

It is known that retinal microglia can phagocytose fluorescent dye released by dying RGCs. To confirm whether microglia could take up Alexa Fluor 488-dextran, retinas from retrograde-labeled eyes that had undergone optic nerve transection were stained with rhodamine-labeled GSL-I. Retinas were rapidly removed in situ and then fixed for 1 hour in freshly prepared 4% paraformaldehyde. The retinas were dissected, washed with phosphate-buffered saline (PBS), and permeabilized in 0.2% Triton X-100 for 15 minutes. After another PBS wash, the retinas were stained with GSL-I (1:200) for 2 hours to label microglia. The density of RGCs was determined by counting Alexa Fluor 488-dextran–labeled cells in three areas per retinal quadrant at three different eccentricities of the retinal radius, for a total of 12 regions per retina. The fraction of Alexa Fluor 488-dextran–labeled cells that were also GSL-I-positive (i.e., microglia that had phagocytosed Alexa Fluor 488-dextran) was calculated from three retinas 14 days after optic nerve transection. 

Histologic Examination

RNFL thickness was measured in retinas from eyes enucleated 7 days after optic nerve transection. Eyes were fixed in 1% paraformaldehyde, 3% sucrose, and 2.5% glutaraldehyde in 0.1 M phosphate-buffered saline. Serial paraffin sections (1-μm thick) crossing the ONH along the vertical meridian of the globe were obtained. After they were stained with toluidine blue, the retinal sections in each eye were observed with a microscope (Axioscope A1; Carl Zeiss, Oberkochen, Germany) attached to a digital imaging rig and were measured. RNFL thicknesses of 400-μm lengths of retina beginning 250 μm from each side of the optic disc on both sides were measured in four consecutive thin sections (Photoshop; Adobe, San Jose, CA). The mean RNFL thickness in each eye was calculated from eight areas of four sections and then compared between transected and nontransected eyes. 

Statistical Analysis

Results are presented as mean ± SD. CMFDA-labeled axonal density (n = 6) and Alexa Fluor 488-dextran–labeled RGC soma density over time (n = 5) were statistically analyzed by ANOVA, followed by Tukey-Kramer testing. Unpaired t-tests were used for the comparison of CMFDA axonal density between transected and nontransected eyes (n = 6). Wilcoxon signed-rank test was used for the comparison of histologic RNFL thickness between transected and nontransected eyes (n = 3). P < 0.05 was considered statistically significant. 

Results

Labeling of Retinal Ganglion Cell Bundles with CMFDA

Intense labeling of RGC axonal bundles with CMFDA could be visualized using the fluorescein filter channel on the CSLO. Indirect ophthalmoscopy demonstrated no visible retinal damage at the site of injection. The labeled RGC axonal bundles were all within one retinal quadrant, were continuous between the injection site and the ONH, and were sparsely spread out. CSLO images correlated with images of whole mounted retinas by conventional fluorescence microscopy (Figs. 1B, 1C). To verify that a rapid intravitreal instillation of fluid did not damage the retina at the injection site, very slow injections using a syringe pump (1 μL/min) were also performed, with identical results (data not shown). To test whether labeling was caused by retinal injury from the dimethyl sulfoxide (DMSO) in which the CMFDA was initially dissolved, DMSO at the same concentration used for diluting CMFDA was injected inferiorly, followed 30 minutes later by injection of CMFDA superiorly. Only superior retinal axon bundles were labeled, indicating that DMSO alone did not prime axons for labeling by CMFDA. 

Figure 1.

 

Representative in vivo images of axon bundles stained with CMFDA and visualized by CSLO. (A) Field 1 (F1) was imaged with the optic nerve head at the center of the image. To calculate fluorescence intensity, F1 was horizontally divided into halves through the center of the optic nerve head. (B) Field 2 (F2) was imaged with the optic nerve head at the edge of the image. (C) Area corresponding to the inset in (B) examined by fluorescence microscopy of the retinal whole mount. (D) Field 3 (F3) was imaged in the peripheral retina. Yellow line: cross-section through which the number of axon bundles was counted. Arrowheads: vessel bifurcations used as fiduciary points. (E) Field 4 (F4) was imaged in the area between F2 and F3. Asterisk: optic nerve head. Scale bar, 100 μm.

To determine the kinetics of axon labeling with CMFDA, CSLO images were taken every 2 hours after injection. For the first 6 hours, axon bundles could not be discerned because of diffuse vitreous fluorescence. However, whole mounted retinas from 1 hour after injection already demonstrated faint CMFDA labeling when examined by conventional microscopy (Fig. 2A). Bundles first became visible by CSLO 8 hours after injection (Fig. 2B) and could be clearly observed 1 day after injection (Fig. 2C). Dye was retained in axon bundles for 1 week but gradually decreased over time in control (nontransected) eyes (Fig. 3). 

Figure 2.

 

Kinetics of axon labeling with CMFDA. (A) Retinal whole mount 1 hour after CMFDA injection demonstrates faint axon bundle labeling by fluorescence microscopy. (B) Axon bundles first become visible by CSLO 8 hours after injection because dye within the vitreous interferes with imaging of the retina at earlier time points. (C) Axon bundles labeled with CMFDA are clearly visible 24 hours after injection. Asterisk: optic nerve head. Scale bar, 50 μm.

Figure 3.

 

CMFDA fluorescence within axon bundles decreases over time in the posterior pole. (A–C) Imaging of CMFDA-stained axons in field 1 of retinas after optic nerve transection. (A) 4 days, (B) 7 days, and (C) 10 days after optic nerve transection. (D–F) Imaging over time of CMFDA-stained axons in control (nontransected) retinas simultaneously with the transected retinas.

Time Course of Axonal Loss after Optic Nerve Transection

To examine early axonal changes after axotomy, CMFDA was intravitreally injected 1 day before optic nerve transection and longitudinally imaged in vivo for up to 5 days. CMFDA axonal counts were normalized to values measured immediately before transection. The intensity of CMFDA axonal bundle fluorescence in transected and control (nontransected) eyes decreased by 31% ± 22% and 30% ± 8% at 4 days after transection, respectively, whereas the number of CMFDA-positive bundles decreased by 46% ± 20% and 47% ± 9%, respectively (Fig. 4). These results suggest that retrobulbar optic nerve transection does not cause significant axonal loss in the retinal nerve fiber layer within the first 4 days. Five days after transection, CMFDA fluorescence in transected and control (nontransected) eyes nonsignificantly decreased by 47% ± 17% and 40% ± 10%, respectively (P = 0.3), whereas the number of CMFDA-positive bundles decreased by 55% ± 16% and 49% ± 6%, respectively (P = 0.27). 

Figure 4.

 

Optic nerve transection does not affect axon bundle fluorescence intensity or number of axon bundles in the first 5 days after axotomy. (A) Change in fluorescence intensity of CMFDA in axon bundles over time. (B) Change in number of fluorescent axon bundles over time. The x-axis reflects days after CMFDA injection or optic nerve transection (ONT). Fluorescence intensity and number of axon bundles were normalized to values obtained at the first day after CMFDA injection. Error bars indicate SD (n = 3). There was no significant difference between control nontransected (dotted lines) and transected (solid lines) eyes at any time point.

Although there was little change in RGC nerve fiber layer axons in the first 4 days after retrobulbar optic nerve transection, this was not true at later time points. To image axons at longer durations than 5 days after transection, it was necessary to instill CMFDA at a later time point because the fluorescence in nontransected eyes typically becomes difficult to image by 7 days after injection. Therefore, we intravitreally injected CMFDA in a separate group of animals 3 days after optic nerve transection and longitudinally imaged in vivo for up to 10 days (Fig. 3). We normalized all measurements to 4 days after transection, given that there was no difference in axon bundle fluorescence or counts at that time point (Fig. 4). 

We found a significant decrease in the intensity of CMFDA fluorescence in axon bundles and the number of CMFDA-positive axon bundles in transected eyes over time (Fig. 5; P < 0.001 by ANOVA) in the peripapillary region (F1), with the greatest decrease occurring between 5 and 7 days (P < 0.001 by Tukey-Kramer). Optic nerve transection produced a rapid reduction in RGC axonal bundle labeling, with loss of 68% ± 29% of the fluorescence 7 days after transection compared with loss of 25% ± 21% in nontransected eyes (P = 0.037). Ten days after transection there was loss of 87% ± 15% of the fluorescence compared with 42% ± 26% in nontransected eyes (P = 0.01; Fig. 5A). The number of labeled RGC axonal bundles decreased by 61% ± 28% and 86% ± 13% at 7 and 10 days after optic nerve transection compared with 26% ± 9% (P = 0.016) and 51% ± 13% (P = 0.005) in nontransected eyes (Fig. 5B). 

Figure 5.

 

Optic nerve transection reduces axon bundle fluorescence intensity and number of axon bundles between 5 and 7 days after axotomy. (A) Change in fluorescence intensity of CMFDA in axon bundles over time. (B) Change in number of fluorescent axon bundles over time. The x-axis reflects days after CMFDA injection or optic nerve transection (ONT). Fluorescence intensity and number of axon bundles were normalized to values obtained at the first day after CMFDA injection. Error bars indicate SD (n = 6). *P < 0.05 (unpaired t-test). **P < 0.05 (ANOVA with Tukey-Kramer).

The degeneration of CMFDA-labeled axon bundles coincided with the loss of the striations within the RNFL in red-free CSLO images (Fig. 6). Individual RGCs labeled with DiR were easily visible with the long-wavelength filter of the CSLO. The number of DiR-positive RGCs decreased in concert with loss of the CMFDA signal. 

Figure 6.

 

Degeneration of CMFDA-labeled axon bundles correlates with loss of striations within the RNFL in red-free CSLO imaging and loss of DiR-positive retrograde-labeled retinal ganglion cell bodies. (A, D) Axon bundles labeled with CMFDA. (B, E) Red-free imaging of RNFL at same retinal location. (C, F) RGC somas retrograde-labeled with DiR at same retinal location. (A–C) 4 days after optic nerve transection. (D–F) 10 days after optic nerve transection.

Loss of Fluorescence Not due to Axonal Leakage

Retrobulbar axotomy could decrease the CMFDA signal either because of a decrease in the number of axons or an increase in the leakage of CMFDA from intact axons. To distinguish these two possibilities, a group of animals was injected with CMFDA 2 weeks (Fig. 7A) and 3 weeks (Fig. 7B) after transection. There were far fewer axon bundles labeled compared with the number labeled 4 days after transection, consistent with the expected loss of RNFL axons 2 or 3 weeks after optic nerve injury. These results suggest that the decrease in CMFDA over time after axotomy results from a decrease in the number of axons and not leakage of CMFDA from preserved axons. 

Figure 7.

 

Few axon bundles are present and capable of being labeled with CMFDA at (A) 2 weeks and (B) 3 weeks after transection. Imaging performed 1 day after intravitreal injection of CMFDA. Asterisk: optic nerve head.

Loss of Fluorescence Not due to Light Exposure

Another contributor to the loss of CMFDA signal could be photic bleaching. To test whether light exposure affects the intensity of CMFDA fluorescence, animals received an intravitreal CMFDA injection. The next day the right eye was imaged and then occluded by suturing the lids with 6-0 silk. The suture was maintained for 7 days and then removed. Mean (± SD) fluorescence intensities 7 days after injection were 55% ± 8% and 59% ± 4% of the fluorescence measured 1 day after injection in the sutured eye and nonsutured eye, respectively (P = 0.547; n = 3). Proportions of axon bundles 7 days after injection were 61% ± 8% and 62% ± 8% in the sutured and nonsutured eye (P = 0.866; n = 3). These results indicate that photic bleaching was not responsible for loss of axonal CMFDA labeling. 

Histologically Measured RNFL Thickness

The RNFL, spanning the inner limiting membrane and the ganglion cell layer, was measured in 1-μm retinal sections stained with toluidine blue. RNFL thickness in eyes 7 days after transection was 24.7 ± 3.4 μm (n = 3). In nontransected eyes, the thickness at the same location was 36.3 ± 2.4 μm (n = 3). This corresponded to a reduction of 38% in the RNFL thickness at 7 days (P < 0.001). 

Longitudinal Change in RGC Soma Counts

In vivo imaging allows serial quantification of actual RGC numbers over time, based on the ability to identify individual RGC somas retrograde-labeled with Alexa Fluor 488-dextran. Relative numbers of fluorescent RGC somas in defined areas of the retina were quantitated by serial imaging before and after optic nerve transection. A decrease in the number of RGCs was observed (Figs. 8A–D). In the first 5 days, the number of RGC somas decreased from 100% to 83% ± 12% (P = 0.023). The most rapid decrease in the number of RGC somas was between 5 and 7 days after optic nerve transection (83% ± 12% to 47% ± 16%; P < 0.001; n = 5). The proportion surviving at 14 days was 24% ± 10% (P < 0.001; Fig. 8E). To assess the possibility that microglial phagocytosis of labeled RGCs was confounding the RGC soma counts, retinas were poststained with rhodamine-labeled GSL-I, a lectin that binds to microglia. Fourteen days after optic nerve transection, 3.3% ± 1.2% of the cells that were positive for Alexa Fluor 488-dextran were also positive for GSL-I (Figs. 9A–D). In nonaxotomized retinas, Alexa Fluor 488-dextran–positive cells were preserved, and few GSL-I–positive cells were observed (Figs. 9D–F). 

Figure 8.

 

Longitudinal evaluation of density of RGCs retrograde-labeled with Alexa Fluor 488-dextran. (A) Overview image of retina by automatic real-time composite mode on CSLO, before optic nerve transection. (B) Magnified view, before optic nerve transection. (C) Corresponding field, 7 days after transection. (D) Corresponding field, 14 days after transection. (E) Quantitation of relative number of labeled RGCs in defined areas of the retina (n = 5). Single asterisk: optic nerve head. **P < 0.05.

Figure 9.

 

Alexa Fluor 488-dextran is infrequently phagocytosed by microglia. Representative photomicrographs from whole-mounted retinas stained with rhodamine-labeled GSL-I from (A–C) transected and (D–F) nontransected eyes. Images were taken focused on the ganglion cell layer, containing GSL-I–stained microglia and capillaries. (A, D) Alexa Fluor 488-dextran. (B, E) GSL-I. (C, F) Merge of (A, D) and (B, E). Arrows: colocalization of GSL-I and Alexa Fluor 488-dextran. Scale bar, 50 μm.

Discussion

To our knowledge these are the first longitudinal observations of changes in a specifically labeled subset of RGC axons in living animals. We used this methodology to show that axonal degeneration and somal degeneration occur in parallel after intraorbital optic nerve transection. The intravitreal injection of CMFDA is simple, not susceptible to transcriptional silencing as are labels dependent on Thy-1 or Brn-3b, unassociated with complications such as cataract, vitreous opacity, or infection, and compatible with daily observation of axonal changes over time in disease or therapeutic models. 

Our results using in vivo CMFDA labeling of axons and Alexa Fluor 488-dextran labeling of somas demonstrated concurrent degeneration of axon and somas at 5 to 7 days after optic nerve transection in the rat. Before that time there were no significant differences in CFMDA axons between transected and nontransected eyes up to 4 days after transection. At 5 days there was a small but nonsignificant decrease in axon bundle fluorescence and counts and a small but significant decrease in RGC soma counts. In general, CMFDA quantification appears to have a higher coefficient of variation than counting the number of retrograde-labeled RGC somas, possibly explaining why the small reduction at 5 days failed to reach statistical significance in the former. Between 5 and 7 days after transection, the parallel reduction in both RGC somas and axons was striking. 

In contrast, Kawaguchi et al. ¹² longitudinally quantitated in vivo the thickness of the rat RNFL after crush with CSLO, finding preservation at 7 days but thinning at 14 days. These findings were confirmed histologically and in a subsequent study using time-domain OCT. ¹³ There are at least three possible reasons for the shorter time (7 vs. 14 days) to detectable axon loss in our study. First, we transected the optic nerve and all RGC axons at once, whereas Kawaguchi et al. ¹² performed a crush with a 60-g vascular clip, which might have caused partial axonal injury. Second, we identified individual axon bundles whereas they quantitated the full extent of the RNFL, which included dead but not yet cleared axons. Third, we monitored dye within the axon, which might have left a dying axon before it underwent dissolution. 

The vitreous is diffusely fluorescent for the first 24 hours after CMFDA injection, interfering with imaging of the retina by CSLO. Although slight axon bundle staining could be observed by CSLO 8 hours after injection, studies of whole mounted retinas revealed that axons were labeled as early as 1 hour after injection. In practice, clear imaging of labeled axon bundles in the RNFL is best performed starting at 24 hours after injection, and we used this time point for normalizing subsequent measurements. 

There are alternative methods for evaluating intraretinal RGC axons in vivo. They are visible with fluorescent imaging when lipophilic carbocyanine dyes such as 4-Di-10-Asp are used for retrograde labeling from the superior colliculus. However, this method stains most of the RNFL and does not allow observation of individual axon bundles over time. Fluorescent protein expression under the control of the Thy-1 promoter also labels RGC axons but is susceptible to transcriptional downregulation as a result of axotomy. ¹⁵ The overall thickness of the RNFL can be estimated by focusing successively on its anterior and posterior borders with a CSLO, but this method is subjective and subject to variability. ¹² It also is unable to identify changes in individual RGC axon bundles. OCT, which is widely used in clinical settings, can be used to calculate RNFL thickness. However, the RNFL thickness measured with OCT is not representative of changes in RGC axons alone because it includes other RNFL components, such as astrocytes and small blood vessels. ¹⁶  

The mechanism by which the vital dye CMFDA stains axons is unclear. Because it is relatively nonpolar, CMFDA is able to freely pass through intact cell membranes and, once inside the cell, is hydrolyzed to a polar and relatively cell-impermeant fluorescent thioether adduct. This dye contains a chloromethyl group that reacts with cellular thiols, ¹⁷ increasing retention over time. However, why the dye labels a sparse subset of axons is unclear. The most convincing explanation is that CMFDA spontaneously enters the RNFL through the inner limiting membrane at the site of intravitreal injection. We confirmed by indirect ophthalmoscopy that CMFDA was injected into the vitreous without visible retina damage, although the actual site of injection could not be imaged by CSLO. Another possibility is that the rapidly injected stream of fluid into the vitreous produced subclinical damage to the superficial retina adjacent to the injection site. However, this is unlikely because axonal staining occurred even when a syringe pump was used to slowly inject CMFDA at 1 μL/min. A third possibility is that axonal damage occurred from toxicity of the DMSO used as diluent for CMFDA before dilution in balanced salt solution. However, the numbers of axonal bundles stained and their intensity was the same when CMFDA was injected at an estimated final vitreal concentration of 6.7% DMSO, compared with 0.3% in other experiments. Furthermore, intravitreal injection of DMSO alone, followed by injection of CMFDA in DMSO at a different site, did not result in staining of bundles at the first site. The focal staining was unlikely to be a concentration-dependent phenomenon (i.e., related to a high local concentration of CMFDA near the site of injection) because increasing the concentration by as much as a factor of 3 does not increase the number of stained axons. The latter would be expected if staining required a threshold for CMFDA concentration at the axon. It is, therefore, unclear why CMFDA stains only bundles of axons in one region and not the entire retina. Further experiments will address this issue. 

There are some drawbacks to staining RGC axons with intravitreal CMFDA. The strength of the signal gradually decreases in healthy axons. The results of the lid suture experiments suggest that the fading of CMFDA is not related to light exposure, but there could be other explanations, such as diffusion or breakdown of dye over time. CMFDA was able to stain only regional bundles of axons; therefore, this technique may be difficult to apply to optic neuropathies in which there is regional injury, such as glaucoma. In vivo imaging of CMFDA-stained axon bundles with CSLO does not have the resolution to identify individual axons. Methods using two-photon microscopy are feasible only at shorter working distances than the axial length of the globe, such as with in vivo imaging of cerebral cortex or spinal cord. ^18–20  

RGC soma density was determined by visualizing Alexa Fluor 488-dextran–labeled cells by CSLO. Lipophilic fluorescent carbocyanine dyes such as DiA, DiI, and DiR are frequently used for retrograde RGC labeling from the superior colliculus. These dyes are long-lasting but are also present in microglia that phagocytose degenerated RGCs. ²¹ Therefore, it became necessary to avoid counting cells with the morphologic features of microglia, a laborious and inexact task, especially with in vivo imaging. In contrast, the fluorescent dextran we used was less phagocytosed by microglia. This difference in microglial phagocytosis could be related to the high molecular weight of the dextran we used (10 kDa), which was approximately 10 times higher than the carbocyanine dyes. Dextrans are hydrophilic polysaccharides and water soluble, and these biophysical characteristics might also decrease their phagocytosis by microglia. Although fluorescent dextrans appeared to take longer to reach RGCs than lipophilic dyes, their use decreased confusion with microglia and thus was advantageous in evaluating RGC density in the retina. Dextrans labeled with infrared-excitable dyes are not commercially available; thus, we could not use the CSLO infrared laser to simultaneously identify RGC somas and CMFDA-labeled RGC axons in the same animals. Instead, we studied parallel groups of animals for axonal and somal degeneration. CSLO and a fundus camera have been used to longitudinally observe axotomized RGCs in vivo in transgenic mice in which cyan fluorescent protein or yellow fluorescent protein is driven by the Thy-1 promoter. ^10,11,22 Leung et al. ¹¹ demonstrated that RGC soma density decreased within 2 weeks after optic nerve crush, similar to our findings using retrograde labeling. One potential confounding factor is that Thy-1 expression itself decreases after RGC axonal damage because of transcriptional silencing. ¹⁵  

Ascertaining where the initial injury occurs in neurodegenerative disorders is essential for understanding their pathogenesis and targeting early therapies. To preserve function, such therapies should target not only somal survival but also axonal preservation. Indeed, neuroprotection of RGC bodies by neurotrophins fails to prevent optic nerve and axonal degeneration. ²³ In vivo longitudinal visualization of individual RGC axon bundles with CMFDA may provide meaningful information on the pathophysiology of optic neuropathies and lead to novel strategies for RGC degenerative disease. 

Footnotes

 Supported by Canadian Institutes for Health Research, Canadian Foundation for Innovation, Canadian Research Chairs program, and Fonds de recherche en ophtalmologie de l'Université de Montréal.

Footnotes

 Disclosure: A. Kanamori, None; M.-M. Catrinescu, None; M. Traistaru, None; R. Beaubien, None; L.A. Levin, None

The authors thank Christopher J. Lieven for technical assistance.