Fundus Autofluorescence and Photoreceptor Cell Rosettes in Mouse Models

Abstract

Purpose.: This study was conducted to study correlations among fundus autofluorescence (AF), RPE lipofuscin accumulation, and photoreceptor cell degeneration and to investigate the structural basis of fundus AF spots.

Methods.: Fundus AF images (55° lens; 488-nm excitation) and spectral-domain optical coherence tomography (SD-OCT) scans were acquired in pigmented Rdh8^−/−/Abca4^−/− mice (ages 1–9 months) with a confocal scanning laser ophthalmoscope (cSLO). For quantitative fundus AF (qAF), gray levels (GLs) were calibrated to an internal fluorescence reference. Retinal bisretinoids were measured by quantitative HPLC. Histometric analysis of outer nuclear layer (ONL) thicknesses was performed, and cryostat sections of retina were examined by fluorescence microscopy.

Results.: Quantified A2E and qAF intensities increased until age 4 months in the Rdh8^−/−/Abca4^−/− mice. The A2E levels declined after 4 months of age, but qAF intensity values continued to rise. The decline in A2E levels in the Rdh8^−/−/Abca4^−/− mice paralleled reduced photoreceptor cell viability as reflected in ONL thinning. Hyperautofluorescent puncta in fundus AF images corresponded to photoreceptor cell rosettes in SD-OCT images and histological sections stained with hematoxylin and eosin. The inner segment/outer segment–containing core of the rosette emitted an autofluorescence detected by fluorescence microscopy.

Conclusions.: When neural retina is disordered, AF from photoreceptor cells can contribute to noninvasive fundus AF images. Hyperautofluorescent puncta in fundus AF images are attributable, in at least some cases, to photoreceptor cell rosettes.

Introduction

The disabling of reactive all-trans-retinaldehyde after its release from photoisomerized visual pigment depends, in part, on its transfer to the cytosolic compartment of photoreceptor outer segment discs by the ATP-binding cassette (ABC) transporter ABCA4, ^1–4 followed by nicotinamide adenine dinucleotide phosphate (NADPH)-dependent reduction to all-trans-retinol by retinol dehydrogenase (RDH) activity. ^5,6 Several RDH enzymes have been identified in photoreceptor cells, including retinol dehydrogenase 8 (RDH8), which is located within outer segments, ⁷ and retinol dehydrogenase 12 (RDH12), which can reduce all-trans-retinal in inner segments. ⁸ Thus far, mutations in RDH8 have not been shown to confer retinal disease in humans; the phenotype in mice is characterized by delayed clearance of all-trans-retinal and modestly enhanced formation of the all-trans-retinal adduct A2E. ^9,10 In humans RDH12 mutations are associated with a severe autosomal recessive retinal dystrophy characterized by severe macular atrophy. ^11,12 Although Rdh12^−/− mice accumulate levels of A2E that are less pronounced than in Rdh8^−/− mice, ^10,13 the Rdh12^−/− mice are more susceptible to light-induced photoreceptor cell degeneration. ¹⁴  

As noted above, an important consequence of inadequate handling of all-trans-retinal is the formation of all-trans-retinal-adducts that accumulate with age as bisretinoid lipofuscin in RPE cells. The bisretinoid of RPE lipofuscin is not produced in RPE; rather it forms in photoreceptor cells. The availability of vitamin A aldehyde for reaction with phosphatidylethanolamine drives the formation of the bisretinoids such as A2E; these fluorophores eventually accumulate as lipofuscin in the RPE. This pathway is reflected in the Abca4^−/− (ATP-binding cassette transporter 4 null mutant) mouse, wherein escalating levels of bisretinoid are present because translocation of retinaldehyde across the photoreceptor disc membrane is inadequate to clear these reactive aldehydes. ^1–3 The accumulation of RPE bisretinoids is accompanied by photoreceptor cell degeneration. ¹⁵ These mechanisms are further exemplified in the Rdh8^−/−Abca4^−/− double knockout mouse wherein the deficiency in Abca4 is augmented by the absence of Rdh8 and bisretinoid levels are elevated as compared with the single knockout Abca4^−/− mice. ¹⁰  

In histological material, the retinae of Rdh8^−/−/Abca4^−/− mice present with photoreceptor cell rosettes, deformations in which photoreceptor cells become circularly arranged with their inner and outer segments directed centrally. ¹⁶ Rosette-like formations within the degenerating retina of other animal models of retinal degeneration also have been described ^17,18 ; included in this group are mice homozygous for the retinal degeneration 8 (rd8) mutation in Crb1. ¹⁹ Based on work using a mouse model of retinal detachment, we recently presented evidence indicating that histologically visible rosettes may correspond to autofluorescent puncta observed in flat-mounted retina and by confocal fluorescence scanning laser ophthalmoscopy. ²⁰ Other mouse models of retinal degeneration also exhibit autofluorescence (AF) spots when imaged in flat-mounted retina or by scanning laser ophthalmoscopy (SLO). ^21–23 However, correspondence between autofluorescent puncta and photoreceptor cell rosettes has not always been demonstrated and the source of the AF has remained puzzling. ²⁴  

Here we have used Rdh8^−/−/Abca4^−/− mice to address issues related to the formation, detection, and impact of RPE lipofuscin, the latter being the principal source of the inherent emission that is imaged noninvasively by confocal SLO (cSLO). Accordingly, we investigated the correspondence between fundus AF and the age-related accumulation of RPE lipofuscin and the temporal correlations between fundus AF and photoreceptor cell degeneration. We also examined the structural underpinnings of autofluorescent fundus spots by probing for correspondence among images acquired as fundus AF, spectral-domain optical coherence tomography (SD-OCT), and light and fluorescence microscopy. 

Methods

Mice and Genotyping

Albino Abca4/Abcr null mutant mice (Abca4^{−/ −}), ²⁵ and pigmented Rdh8^−/−Abca4^{−/ −} mice (gift from Krzysztof Palczewski, Case Western Reserve University, Cleveland, OH, USA) were homozygous for Rpe65-Leu450. Wild-type mice (C57BL/6N, Charles River Laboratory, Wilmington, MA, USA; C57BL/6J, Jackson Laboratory, Bar Harbor, ME, USA) were Rpe65-450Met. The Rdh8^−/−Abca4^{−/ −} and C57BL/6N mice were homozygous for the rd8 mutation in Crb1. ²⁶ Animals were housed under 12-hour on-off cyclic lighting with in-cage illuminance of 30 to 80 lux or 10 lux (C57BL/6N). After mice were euthanized and after enucleation, murine eyes were extracted for HPLC analysis or were processed for histology. Mouse DNA was PCR-amplified with primers to detect the following: targeted deletion of Abca4, 5′-CCACAGCACACATCAGCATTTCTCC-3′ (forward) and 5′-TGCGAGGCCAGAGGCCACTTGTGTAGC-3′ (reverse) (455-bp product indicated targeted deletion); Rpe65 Leu450Met variant, 5′-ACCAGAAATTTGGAGGGAAAC-3 (forward) and 5′-CCCTTCCATTCAGAGCTTCA-3 (reverse) (180- and 365-bp bands after MwoI restriction enzyme digestion indicated the leucine variant) ^25,27 ; Rdh8 null allele, 5′-tccgccttggaaacctgagccagaag-3′ (forward) and 5′-tgcgaggccagaggccacttgtgtagc-3′ (reverse) ⁹ ; rd8 mutation in Crb1, 5′-GTGAAGACAGCTACAGTTCTGATC (forward 1) and 5′-GCCCCTGTTTGCATGGAGGAAACTTGGAAGACAGCTACAGTTCTTCTG (forward 2) and 5′-GCCCCATTTGCACACTGATGAC (reverse) (450-bp product indicated the rd8 mutation). ¹⁹  

The research was approved by the Institutional Animal Care and Use Committee and was performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Visual Research. 

Acquisition and Analysis of Fundus AF SD-OCT Images

Mice were anesthetized with an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg) and pupils were dilated with 1% tropicamide and 2.5% phenylephrine (Akorn, Inc., Lake Forest, IL, USA). Corneal desiccation was reduced by topical application of GenTeal Liquid Gel (Novartis, East Hanover, NJ, USA). ¹⁵ The mouse was positioned on a custom-made platform and body temperature was maintained with a heating blanket, temperature controller (Model TC-1000), and thermistor probe (YSI-451; IITC Life Science, Woodland Hills, CA, USA). 

Fundus AF images (55° wide-field lens; 488-nm excitation; 0.98-mm detection pupil) were obtained with a cSLO (Spectralis HRA; Heidelberg Engineering, Heidelberg, Germany) with laser power set at approximately 280 μW and sensitivity at 95 to 100 after visual pigment was bleached for 20 seconds. Nine successive frames were acquired with the high-speed mode and frames were saved in non-normalized mode. 

To quantify fundus autofluorescence (quantitative fundus AF [qAF]) the Spectralis was equipped with an internal fluorescence reference ²⁸ to account for variations in laser power and detector sensitivity and focus was adjusted as previously published. ¹⁵ Using a dedicated image analysis program written in IGOR (Wavemetrics, Lake Oswego, OR, USA), mean gray levels (GL) were calculated from eight predefined segments around the optic disc (Fig. 1), blood vessels were excluded by histogram analysis, and qAF was calculated by normalization to the GL of the reference after subtraction of zero light (GL₀) and inclusion of a reference calibration factor. ^15,28  

Figure 1

 

Quantitation by HPLC of bisretinoid lipofuscin and qAF in Rdh8^−/−Abca4^−/− and Abca4^−/− mice. (A) The A2E quantified in Rdh8^−/−Abca4^−/− mice by reverse-phase HPLC (left y-axis) and qAF (right y-axis) are plotted as a function of age. The A2E values are mean ± SEM; A2E: four eyes per sample and two to six samples per mean; qAF: 12 to 15 mice at ages 1 to 6 months and 1 mouse at age 9 months. (B) The qAF values in Rdh8^−/−Abca4^−/− mice were calculated as the average of the eight segments outlined in red. The eight segments are located between 8.25° and 19.25° from the center of the optic disc (yellow cross). The gray levels in each segment are normalized to values determined for the internal fluorescent reference (top); sampling area in the reference is indicated by the red rectangle. Superior (S-1–4) and inferior (I-1–4) segments are labeled: A, anterior; P, posterior. Values presented in each segment are qAF ± SEM (right eyes of eight mice, age 4 months). (C). High-performance liquid chromatography quantitation of A2E in Rdh8^−/−Abca4^−/− and Abca4^−/− mice are normalized to peak A2E levels (Rdh8^−/−Abca4^−/− : 4 months; Abca4^−/− 8 months) and plotted as a function of age. Normalized mean ± SEM; four eyes per sample and two to six (Rdh8^−/−Abca4^−/− ) or two to five (Abca4^−/− ) samples per mean.

In fundus AF images, autofluorescent puncta were measured using the Spectralis image analysis tools. The transverse resolution of the laser scanning images is approximately 10 μm. ²⁹  

Horizontal SD-OCT scans (1.8 mm) were also acquired with the Spectralis in high-resolution mode with averaging of 100 frames. A corneal radius of curvature of 1.22 was used after calibrating retinal thickness measurements in the SD-OCT scans to thicknesses at corresponding positions (0.2-mm intervals) in histological sections through the optic nerve head of four mice; the value used was consistent with previous determinations in mice. ^30–34 Because axial length influences image magnification, distances from the base of the optic nerve to anterior cornea were measured in paraffin sections (10-μm thickness; ×20 magnification) at 3 and 6 months of age using Photoshop version 7 (Adobe Photosystems, Inc., San Jose, CA, USA). Measurements obtained in C57BL/6J (2.586 ± 0.034, mean ± SEM) and Rdh8^−/−Abca4^−/− (2.574 ± 0.018, mean ± SEM) mice (four eyes per group) were not significantly different (one-way ANOVA, P > 0.05). 

Histometric Analysis

Mouse eyes were fixed by immersion in 4% paraformaldehyde, 16.8% isopropyl alcohol, 2% trichloroacetic acid, and 2% zinc chloride in phosphate buffer for 24 hours at 4°C after marking the superior pole. Whole eyecups were embedded in paraffin and sectioned at a thickness of 8 μm. Sections were then counterstained using hematoxylin and eosin (H&E). Morphologic observations and light microscopy were performed using a digital imaging system (Leica Application Suite; Leica Microsystems, Welzlar, Germany). Outer nuclear layer (ONL) thickness was measured at 200-μm intervals superior and inferior to the edge of the optic nerve head (ONH) along the vertical meridian. ²⁷ The section most centrally located within the ONH was used for measurements after imaging with a digital imaging system (×20 objective; Leica Microsystems); composite images were then created in Photoshop CS5. Serial H&E-stained sagittal sections were also collected from entire mouse eyes and stained to examine for structure changes and rearrangement of photoreceptor cells. The ONL area was also calculated as the sum of the ONL thicknesses in superior and inferior retina (0.2–2.2 mm) and multiplied by the measurement interval of 0.2 mm. 

Additionally, fixed eyes (1, 4, and 6 months) were mounted in Tissue-Tek OCT compound (Sakura Finetek, Torrance, CA, USA) and sectioned horizontally in a cryostat (15 μm). Sections through the ONH were stained with 4′,6-diamidino-2-phenylindole, dihydrochloride (DAPI) to label nuclei and were examined under a fluorescence microscope (×40 objective; Zeiss Axioplan II microscope with AxioCam HRc camera; Carl Zeiss, Thornwood, NY, USA) with DAPI (359/461 nm, excitation/emission maxima) and fluorescein (490/525) appropriate filter sets. The RPE lipofuscin has a broad yellow/orange emission (500–800 nm) ³⁵ ; with the 525-nm band-pass filter, a narrow spectral window of this emission is captured and the fluorescence is green. 

Quantitative HPLC

The A2E and A2E-isomers were measured in mouse eyes (four eyes per sample) by HPLC (Alliance system; Waters Corp., Milford, MA, USA) ²⁵ after homogenization and extraction in chloroform/methanol. ³⁶ Absorbance peaks were identified by comparison with external standards. Molar quantities per eye were calculated from peak areas using standard concentrations determined spectrophotometrically together with published extinction coefficients, and by normalization to total sample volumes. Mean values for genotype and age were determined by averaging multiple independent samples. 

Statistical Analyses

Analyses were performed using Prism 5 (GraphPad Software, La Jolla, CA, USA) and the statistical tests as indicated. 

Results

Quantitative Fundus AF

Using a previously published approach, ¹⁵ we measured fundus AF levels in Rdh8^−/−/Abca4^−/− mice from 1 to 9 months of age by calculating qAF values averaged over eight predetermined fundus segments (Figs. 1A, 1B). Measurements revealed an age-related increase in qAF, with the progressive increase in qAF being sustained from 1 to 9 months of age (oldest age studied) (Fig. 1B). The qAF coefficient of variation (CV) (SD/mean × 100) ranged from 25.5% (age 3 months) to 33.8% (age 1 month). The mean qAF values for the four superior (1.28 ± 0.11, n = 15; mean ± SEM) and four inferior segments (1.15 ± 0.10, n = 15; mean ± SEM) were normally distributed (D'Agostino and Pearson normality test) but the differences between the means at 1 month (superior: 1.28 ± 0.11; inferior: 1.15 ± 0.11; mean ± SEM; n = 15), 2 months (superior: 1.54 ± 0.10; inferior: 1.56 ± 0.11; mean ± SEM; n = 15), and 3 months (superior: 1.65 ± 0.10; inferior: 1.78 ± 0.13; mean ± SEM; n = 15) were not significant (P > 0.05, unpaired t-test). 

Quantitation of Bisretinoids by HPLC

In eyes obtained from Rdh8^−/−/Abca4^−/− mice, A2E, one of the bisretinoid constituents of RPE lipofuscin, was measured at various ages by integrating HPLC peak areas and normalizing to standard samples of known concentration. For comparison, we plotted the A2E dataset on the left y-axis of Figure 1A, with qAF displayed on the right y-axis. Although qAF exhibited an increasing trend from 1 to 9 months of age, A2E levels increased between 2 and 4 months of age and decreased thereafter. To contrast the profile of A2E accumulation in Rdh8^−/−/Abca4^−/− mice with that in Abca4 ^−/− mice, we also plotted previously published data ¹⁵ for A2E accumulation in Abca4 ^−/− mice along with the data for Rdh8^−/−/Abca4^−/− mice (Fig. 1C). As shown in Figure 1C, in both Abca4 ^−/− and Rdh8^−/−/Abca4^−/− mice, ^25,37 A2E levels increased between 2 and 4 months of age but the rate of increase was greater in the Rdh8^−/−/Abca4^−/− mice (Fig. 1A). Specifically, the slopes (±SEM) of the regression lines for Abca4 ^−/− and Rdh8^−/−/Abca4^−/− mice were 10.3 ± 2.3 (goodness of fit, R ², 0.62) and 27.6 ± 6.2 (goodness of fit, R ², 0.71), respectively. In Rdh8^−/−/Abca4^−/− mice, A2E levels peaked at 4 months of age, whereas in Abca4 ^−/− mice, the maximum occurred at 8 months of age. 

Thickness of ONL

We measured ONL thickness as an indicator of photoreceptor cell viability in Rdh8^−/−/Abca4^−/− mice and plotted thicknesses in 200-μm intervals superior and inferior to the ONH in the vertical plane (Fig. 2A). Differences between Rdh8^−/−/Abca4^−/− and wild-type (C57BL/6N) (age 8 months) mice were readily visible, particularly in the superior hemisphere of retina. Thinning of ONL in Rdh8^−/−/Abca4^−/− mice continued to progress from 3 to 6 months of age (Figs. 2A, 2B). For comparative purposes, we calculated ONL area from the sum of ONL thicknesses in superior and inferior retina (0.2 to 2.2 mm), multiplied by the measurement interval of 0.2 mm. Relative to the wild-type C57BL/6N mice (2.3 × 10⁵ mm² ± 0.01, mean ± SEM) mean ONL area in 2-month-old (2.0 × 10⁵ mm² ± 0.06) and 4-month-old (1.8 × 10⁵ mm² ± 0.03) Rdh8^−/−/Abca4^−/− mice was reduced by 11% and 21%, respectively. At 6 months of age, mean ONL area (1.4 × 10⁵ mm² ± 0.02) was diminished by 40% (P < 0.05; nonparametric Kruskal-Wallis and Dunn's multiple comparison test). In C57BL/6N mice, ONL thicknesses were not different at 3, 6, and 8 months (data not shown). 

Figure 2

 

Outer nuclear layer thinning in Rdh8^−/−/Abca4^−/− mice. (A) Quantification of ONL thickness in Rdh8^−/−/Abca4^−/− (ages 2–6 months) and wild-type (C57BL/6N) mice (8 months), respectively. Means ± SEM are plotted as a function of distance from the ONH in the inferior and superior hemispheres. Numbers of mice are presented in parentheses. (B, C) Light micrographs of retinae of Rdh8^−/−/Abca4^−/− mice at 2 (B) and 6 (C) months of age document progressive ONL thinning.

Autofluorescence in Retinal Laminae

To better understand sources of fundus AF in Rdh8^−/−/Abca4^−/− mice, we also examined cryostat sections of Rdh8^−/−/Abca4^−/− mouse retinae in areas of normal lamination (Fig. 3). An inherent autofluorescence was observed not only in RPE cells, as expected, but also in the lamina occupied by photoreceptor outer segments. This autofluorescence in the Rdh8^−/−/Abca4^−/− mice was more pronounced in older mice. 

Figure 3

 

Fluorescence photomicrographs of mouse retina. Rdh8^−/−Abca4^−/− mouse, age 1 month (A–C), age 4 months (D–F), and age 6 months (G–I). DAPI fluorescence, 7.4-ms exposure; autofluorescence (FITC-filter), 179-ms exposure.

In Vivo and Ex Vivo Imaging of Autofluorescent Puncta

In fundus AF images, the Rdh8^−/−/Abca4^−/− mice exhibited readily detectable AF spots predominantly in the inferior-nasal quadrant of the eye (Figs. 1E, 4, 5A, 5B). Measurements of the AF puncta in fundus images were acquired at 2, 4, and 6 months of age (two mice, 15 puncta per retina at each age). The mean diameter of the AF puncta was 30 μm (±0.96, SEM; range, 20–50 μm; sample size, 90). The data were normally distributed (D'Agostino & Pearson omnibus normality test) and no differences were observed among the age groups (one-way ANOVA, P > 0.05). Puncta of intense AF were only occasionally observed at 1 month of age, were increased in number between 3 and 5 months of age, and had diminished in frequency at 7 months of age (Fig. 4). In C57BL/6J mice, autofluorescent puncta were not observed at these ages (data not shown). 

Figure 4

 

Fundus AF images of Rdh8^−/−Abca4^−/− mice at the ages indicated. Left eyes. (A–D) Autofluorescence puncta emerge first in the inferior–anterior quadrant of the fundus and are also more abundant there.

Figure 5

 

Multimodal imaging of retinae of Rdh8^−/−Abca4^−/− and Abca4^−/− mice. (A, B) Representative fundus AF (488-nm excitation) image of Rdh8^−/−Abca4^−/− mouse (age 4 months; left eye) (A). Higher-magnification fundus AF image of inferior retina (age 5 months; left eye) (B). The horizontal green line with arrow indicates the position of the corresponding horizontal SD-OCT scan. (C) Spectral-domain OCT reveals hyperreflectivity extending through the ONL and IPL-attributable bands; lower image is region of the hyperreflectivity at higher magnification. GC/IPL, ganglion cell and inner plexiform layers; INL, inner nuclear layer; NFL, nerve fiber layer. (D, E) Light microscopic image of Rdh8^−/−/Abca4^−/− mouse retina (age 3 months) stained with H&E. Rosette-like rearrangements of photoreceptor cells extend from ONL and across IPL toward INL. Arrow, photoreceptor outer segments project inward to form the core of the rosette. (F–H) Fluorescence micrographs of Rdh8^−/−/Abca4^−/− retina (20-μm cryostat section). DAPI-stained nuclei (F); white arrow, nuclei of photoreceptor rosette extending from ONL through OPL. Inherently autofluorescent punctum (G) is observed in OPL using fluorescein (490/525) filter set. By merging images (H) of DAPI-stained nuclei (blue) with AF (green), it is clear that the AF is in the center of the rosette. (I) Fundus AF image of Abca4^−/− mouse (age 8 months, right eye). Autofluorescence puncta are observed inferiorly.

Spectral-domain OCT images of Rdh8^−/−/Abca4^−/− mouse retinae were acquired at 2 to 8 months of age (Fig. 5C). In these images, the laminar organization of the retina was often interrupted by areas of abnormal outer retinal hyperreflectivity. These foci extended through the ONL and outer plexiform layer (OPL)-attributable reflectivity bands. 

Histological analyzes were performed using H&E-stained paraffin sections of mouse retina at 1 to 6 months of age. Under the light microscope, the laminar arrangement of retina was often distorted by outer retinal folds and rosette-like structures (Figs. 5D, 5E). These aberrations involved a spherical organization of photoreceptor cell nuclei that usually occupied the ONL, extended across the OPL, and could also encroach on the inner nuclear layer. The center of the rosettes appeared to be occupied by photoreceptor inner and outer segments. Measurement of rosettes (including nuclei and inner core) in the H&E-stained sections (×20 objective; sample size 13) revealed a mean height of 44.1 (± 0.07, SEM) and mean width of 35.6 μm (±0.7, SEM; width at half height; range, 29.2–74.5). 

To observe for the presence of AF in the rosettes, we examined cryostat (horizontal) sections that had been stained with DAPI to label nuclei. As shown in Figures 5F through 5H, circularly arranged DAPI-stained nuclei were observed in the ONL and OPL. The cores of these circular figures were positioned in the OPL and exhibited an AF when viewed by fluorescence microscopy. The dimensions of the rosette (based on DAPI staining; Fig. 5F) were determined using the integrated measurement module of the microscopy (×40 objective) and were found to be 35.2 μm (horizontal) and 28.1 μm (vertical). 

Autofluorescent puncta were found to also be prominent in inferior–anterior fundus of Abca4^−/− mice (Fig. 5I) that do not carry the Crb1^rd8 mutation. Thus, these aberrations are not only conferred by the Crb1^rd8 mutation in Rdh8^−/−/Abca4^−/− mice. 

Discussion

Previous studies have shown that mice carrying a double knockout of Rdh8^−/−/Abca4^−/− exhibit rapid accumulation of the bisretinoids A2E and all-trans-retinal dimer and early-onset photoreceptor cell degeneration. ¹⁶ Our findings are consistent with this earlier work. As compared with single knockout Abca4^−/− (albino) mice at age 4 months, A2E levels in the pigmented Rdh8/Abca4^−/− mice were more than 2-fold greater. Thinning in the ONL in the Rdh8/Abca4^−/− mice also was detected at an earlier age (Abca4^−/− at age 8 months; Rdh8/Abca4^−/− at age 4 months). We cannot compare qAF levels in the Abca4^−/− and Rdh8/Abca4^−/− mice, because the former were albino and the latter were pigmented. Even in the presence of equal levels of RPE lipofuscin, the absence of melanin in the albino mice can be expected to confer higher qAF values due to reflected light. ³⁸  

In Rdh8/Abca4^−/− mice, a decline in levels of the bisretinoid lipofuscin fluorophore A2E coincided with photoreceptor cell loss as measured by ONL thickness. We previously found that A2E also declined in Abca4^−/− mice, concurrent with thinning of ONL. ¹⁵ Yet despite this decrease in bisretinoid, fundus AF continues to increase in both the Abca4^{−/− 15} and Rdh8/Abca4^−/− mice (Fig. 1). We formerly postulated that fundus AF can be potentiated in association with photoreceptor cell dysfunctioning and degeneration. ^39,40 Perhaps, therefore, the continued increase in fundus AF observed after 4 months of age in Rdh8/Abca4^−/− mice originates from the outer segments of degenerating photoreceptor cells. This notion is consistent with the age-related increase in AF observed in outer segments when viewed by fluorescence microscopy (Fig. 3). The identities of the specific fluorophores accounting for this continued increase in qAF are not yet known. Other bisretinoids potentially detectable by fundus AF could include A2PE, the precursor of A2E located in photoreceptor outer segments. ⁴¹ In this regard, two factors may enable the detection of fundus AF derived from photoreceptor cells positioned anterior to the RPE and allow this AF to appear particularly bright in pigmented Rdh8/Abca4^−/− mice. First, the fluorescence emission from outer and inner segments would not be subject to absorbance by melanin as would be the case for lipofuscin in RPE. And second, although the fluorophores in the outer segments and perhaps inner segments of photoreceptor cells would presumably be present at lower concentrations, they also would be less compacted than in RPE cells and thus their fluorescence would undergo less self-absorption. 

The background of the Rdh8^−/−/Abca4^−/− mice we studied was complicated by the presence of the rd8 mutation in Crb1. Nonetheless, in our study of the features of AF in degenerating retina, we were able to use this complex phenotype to advantage. Retinae of both the Rdh8^−/−/Abca4^{−/− 16} (Figs. 4, 5) and Crb1^rd8/Crb1^rd8 mutant mice ¹⁹ are described as having photoreceptor cell rosettes. Interestingly, patients with retinal degeneration due to CRB1 mutations also exhibit intraretinal hyperreflective lesions that the authors suggest could represent the rosettes observed in mice. ⁴²  

Here we visualized photoreceptor cell rosettes in mice by SD-OCT and by H&E and DAPI staining of retina sections. Autofluorescence originating in the center of the rosettes was detected by fluorescence microscopy and as AF puncta visible in vivo with imaging by SLO. The rosettes observed here varied in size. Using the magnification bars provided in the figures, we confirmed that the size range in our study was similar to the range presented in earlier publications. ^23,43,44  

In other mouse models of retinal degeneration, AF spots have been observed in images of flat-mounted retina or by cSLO. This feature is exemplified by Ccl2/Cx3cr1 mice, wherein granular-like hyperautofluorescent lesions in fundus AF images have been shown to correspond to hyperreflective lesions in the ONL-attributable band in OCT scans and to circularly arranged ONL nuclei in histological sections. ²³ In other cases, the AF has been attributed to microglial cells. ^21,22 For instance, CD11c-positive activated microglia have been identified in conjunction with disruptions in the ONL that present as photoreceptor cell rosettes. ¹⁷ Indeed, it has been suggested that in mice, although not in human retina, perivascular and subretinal microglia are the major lipofuscin-producing cells in normal aged mouse retina. ²² However, evidence based on fluorescence spectral profiles, age-associated accumulations, and disease relatedness indicates that fundus AF in mice originates primarily from lipofuscin in RPE cells, as is also the case in humans. For instance, in both humans and mice, RPE lipofuscin ^37,45–48 and fundus AF ^15,38,49,50 increase with age. Fundus AF and RPE lipofuscin both also increase with ABCA4 deficiency in humans ^51,52 and mice. ^53,54 The spectral characteristics of the fluorescence emission generated from RPE lipofuscin in mice also is similar to the fundus AF spectra recorded noninvasively in humans. ^55,56 In a mouse retinal-detachment model, we found that autofluorescent puncta visible in the fundus exhibited emission spectra (emission peaks at 581 nm and 629 nm with excitations at 488 and 561 nm, respectively) that were consistent with an origin from the bisretinoid fluorophores that form in photoreceptor cell outer segments and become the constituents of RPE lipofuscin. ²⁰ Moreover, the red-shift in emission maxima we recorded from the puncta at a longer (561 nm) versus shorter (488 nm) excitation wavelength was characteristic of the emission red-shifts observed for both fundus autofluorescence ³⁵ and RPE lipofuscin. ⁵⁶  

Of what significance are these findings? Fundus AF images obtained in humans presenting with some retinal diseases exhibit autofluorescent puncta. These disorders include central serous chorioretinopathy, ⁵⁷ fundus albipunctatus, ⁵⁸ and rod-cone degeneration caused by homozygous mutations in nuclear receptor transcription factor NR2E3. ²¹ In the latter case, it is suggested that these hyperautofluorescent spots correspond spatially to rosette-like photoreceptor cell rearrangements demonstrable by OCT. Nonetheless, the hyperautofluorescence of the puncta has been attributed to macrophage infiltration. Histopathologic studies of some cases of AMD and autosomal-dominant RP ^24,59,60 have also documented photoreceptor cell rosettes. The lumina of the rosettes in both humans and mice have been shown to contain rhodopsin, ^43,60 consistent with the presence of outer segments. In the rosettes described here, photoreceptor outer segments (OS) also extended into the center of the rosettes (Figs. 4, 5). Hyperreflective ovoid structures referred to as outer retinal tubulations ^61,62 are observed in many forms of human retinal degeneration, including RP, Stargardt disease, and acute zonal occult outer retinopathy; these lesions are likely analogous to photoreceptor rosettes. 

After considering published works together with our current findings, we conclude that autofluorescent puncta, observed in at least some mutant mouse lines, likely correspond to the outer segment–inhabited interior of photoreceptor cell rosettes, an abnormal feature of retinal disorders in both humans and mice. These findings do not preclude the possibility that microglial/macrophage cells are present in association with the rosettes; yet, we suggest that even given this eventuality, the AF is likely derived from outer-segment debris that has accumulated extracellularly or within macrophages after phagocytosis. 

Our interest in undertaking this work was 2-fold. First, we wanted to further our understanding of fundus AF in relation to RPE lipofuscin accumulation and photoreceptor cell degeneration. Second, we aimed to determine whether photoreceptor cell rosettes/tubulations emit an AF. Common to both of these issues is the question of whether photoreceptor cells can contribute to aberrations in fundus AF. Based on our evidence of ONL thinning, the photoreceptor cells in the Rdh8^−/−/Abca4^−/− mice are dysfunctional and progressively nonviable. In addition to the evidence of photoreceptor cell dysfunctioning based on diminishing numbers of nuclei in the ONL, fluorescence microscopy revealed that OS in the Rdh8^−/−/Abca4^−/− mice exhibit a readily detectable AF that increased with age. Due to outer segment turnover, ⁶³ bisretinoids do not normally accumulate in photoreceptor cells; thus we suggest that the increased AF in OS reflects dysfunctioning and degenerating photoreceptor cells. Mishandling of all-trans-retinal due to deficiency in Rdh8 and Abca4 would rapidly amplify bisretinoid formation in photoreceptor cells, leading to increased fundus AF intensity. ³⁹ We also found that the interior of rosettes, a space occupied by photoreceptor inner and outer segments, was associated with an AF that was considerably brighter than the AF of surrounding puncta-free retina. Rosette formation undoubtedly disrupts the normal interface between OS and the apical domain of RPE, a relationship essential to daily photoreceptor maintenance (phagocytosis, retinoid exchange, nutrient exchange). Whether the AF intensity here is simply attributable to an accumulation of unphagocytosed outer segment membrane or whether the photoreceptor cells forming the rosette are subject to accelerated production of bisretinoid, is uncertain. Nonetheless, taken together, these findings illustrate that photoreceptor cells can serve as an abnormal source of heightened fundus AF. These findings also may be relevant to our understanding of the paracentral rings of elevated AF in retinitis pigmentosa ⁶⁴ and/or the bright AF flecks that characterize recessive Stargardt disease. ^51,65  

Acknowledgments

Supported by National Institutes of Health Grants RO1EY12951 and P30EY019007, Foundation Fighting Blindness, and by a grant from Research to Prevent Blindness to the Department of Ophthalmology, Columbia University. Also acknowledged is the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Biomedical Laboratory Research and Development Program through Veterans Administration Merit Award 1I01BX000669-01 (JMS), and an unrestricted grant from Research to Prevent Blindness to the Department of Ophthalmology (Ross Eye Institute), University at Buffalo-SUNY. The contents do not represent the views of the Department of Veterans Affairs or the US government. JMS is a member of the SUNY Eye Institute. 

Disclosure: E. Flynn, None; K. Ueda, None; E. Auran, None; J.M. Sullivan, None; J.R. Sparrow, None