Microglia dynamics in retinitis pigmentosa model: formation of fundus whitening and autofluorescence as an indicator of activity of retinal degeneration

In patients with retinitis pigmentosa (RP), color fundus photography and fundus autofluorescence (FAF) have been used to estimate the disease progression. To understand the origin and the diagnostic interpretation of the fundus color and FAF, we performed in vivo imaging of fundus color and FAF together with histological analyses of the retinal degeneration process using the RP model mice, rd10. FAF partly represented the accumulation of microglia in the photoreceptor outer segments. Fundus whitening suggested the presence of apoptotic cells, which spatiotemporally preceded increase in FAF. We observed two patterns of FAF localization, arcuate and diffuse, each indicating different pattern of apoptosis, wavy and diffuse, respectively. Diffuse pattern of apoptosis was suppressed in dark-raised rd10 mice, in which outer nuclear layer (ONL) loss was significantly suppressed. The occupancy of FAF correlated with the thinning rate of the ONL. Fractalkine, a microglia chemotactic factor, was detected in apoptotic photoreceptors, suggesting chemokine-induced recruitment of microglia into the ONL, which paralleled with accelerated ONL loss and increased FAF occupancy. Thus, we propose that the degree of photoreceptor apoptosis and the rate of ONL thinning in RP patients might be read from the fundus color and the FAF.


Results
Whitish retina and increased autofluorescence was observed in retinitis pigmentosa model: rd10 mice. In wild type (WT) C57BL/6 J ocular fundus, no change in retinal color tone or no hyperfluorescent spot by FAF was observed (Supplemental Fig. 1A). Ramified form of microglia resided in the inner and outer plexiform layers (IPL and OPL) in normal retina (Supplemental Fig. 1B). In the retinas of rd10 mice, the whitish part was observed in color fundus photographs from postnatal 18 to 21 days (P18-P21) followed by multiple granular white dots at P21 (Fig. 1A). Increased FAF became evident from P19 and peaked at P22, which gradually decreased thereafter (Supplemental Fig. 2).
Whitish retinas on color fundus was related to photoreceptor apoptosis. To investigate the relationship between the whitish fundus and the apoptosis of the photoreceptors in rd10 mice retinas, fundus color photographs and distributions of apoptotic cells in isolated flat-mounted retina were compared during the degeneration process. In color fundus photographs, the entire fundus oculi became uniformly white on P19 (Fig. 1A). Similarly, the number of apoptotic cells that were labeled by an antibody against single strand deoxyribonucleic acid (ssDNA) peaked on P19. The occupancy of the whitish fundus and the apoptotic cells correlated well (Fig. 1B, C). The calculated regression line was as follows: (Area ratio of ssDNA + %) = 0.289 (area ratio of whitish fundus %) + 0.432 (R 2 = 0.901, p = 1.61 × 10 -12 ).
To understand the pathological phase of the uniformly whitish fundus, histological examination was carried out on the retinal sections of rd10 retinas at P19. At this time point, no autofluorescence appeared at the center of the fundus by FAF imaging (Fig. 1D). However, a considerable number of amoeboid-type microglia were infiltrating into the ONL and phagocytizing ssDNA labeled apoptotic photoreceptors. Some microglia were further migrating into the depth of the photoreceptor outer segments (OS) (Fig. 1E, F, G).
Widespread diffuse distribution of apoptosis precedes increased FAF and substantial ONL loss in rd10 mice. To investigate the relations among FAF, apoptosis of the photoreceptors and ONL loss, the area with hyperfluorescence in FAF imaging, ssDNA positive cells in flat-mounted retina and ONL thickness in retinal sections were quantified and plotted over time in the central (Fig. 2) and the peripheral (Fig. 3) retinas (n = 4). In the central part of rd10 retinas, widespread diffuse distribution of apoptotic photoreceptors indicated by ssDNA were observed on P18 and peaked on P20, while FAF spots with high intensity were detected from P20 and were most evident on P22 ( Fig. 2A, B). The ONL thickness dramatically decreased by P24, and then the decrease slowed down. The FAF was maximum at the peak of ONL thinning. Altogether it was suggested that photoreceptor apoptosis precedes substantial ONL loss and the peak of FAF occupancy (Fig. 2B).
In the peripheral rd10 retina, a characteristic arcuate apoptosis area indicated by ssDNA was observed on P18, followed by apoptosis in widespread diffuse area on P20 (Fig. 3A, B). By FAF imaging, an arcuate hyperfluorescent area was also observed on P18, followed by the appearance of widespread diffuse hyperfluorescent spots on P22, again suggesting that apoptosis of photoreceptors precede the emergence of hyperfluorescent FAF spots. The ONL thickness reduced more slowly in the peripheral part of the retina compared to the central part (Fig. 3B). The time course of FAF occupancy showed two-peaks that coincided with the peaks in the rate of ONL thinning (Fig. 3B).
Diffuse cell apoptosis, diffuse FAF spots and ONL loss was significantly suppressed in rd10 mice raised in dark. Since it has been known that retinal degeneration is delayed in rd10 mice if raised in dark, we studied how FAF, cell apoptosis and ONL loss are affected by dark environment. When bred in dark, hyper-fluorescence by FAF imaging was not evident in the central retina up to P60. The apoptosis of photore- at the ONL level, labeled with an apoptosis marker ssDNA (white). Scale bars, 100 μm. Fundus started to whiten on P19. Whitening spread throughout the whole surface on P20. On P22, white granular spots on the subretina were observed. Around P30-P60, the color tone of the retina became normal. Remarkable retinal atrophy was observed after P120. (B) Graphs show time course of fundus whitening area (%) (White marker) and ssDNA + area (%) (Red marker). The markers of the graph show mean and standard deviation (SD) (n = 3). Peaks of both parameters coincide. (C) Scatter plot shows the relationship between the ssDNA + area (%) and the whitening area (%). Measurement points were P17, 18, 19, 21, 30,  www.nature.com/scientificreports/ ceptors indicated by ssDNA was also suppressed. The ONLs were significantly thick compared to those bred under regular light cycle (Fig. 4A, B). In the peripheral retina, interestingly, a characteristic arcuate localization of apoptotic cells was observed on P18, but the presence was transient (P18-P20) and no widespread diffuse apoptosis was observed. Likewise, an arcuate hyperfluorescent area was observed on P18 but the emergence of   www.nature.com/scientificreports/ widespread diffuse FAF was mostly suppressed. The ONL thinning in the peripheral retina was less suppressed by dark environment compared to the central retina (Fig. 5A, B). However, unlike those bred under regular light cycle, peripheral FAF and cell apoptosis showed only one peak around P18, which was the time ONL thinning was most rapid (Fig. 5B). We observed two types of FAF in the retina of rd10 mice. Early arcuate hyperfluorescent area was observed on P18 in the peripheral rd10 retina raised either under regular light cycle or in dark (Supplemental Fig. 3A), while widespread diffuse hyperfluorescent spots that peaked around P22 was observed in both central and peripheral rd10 retina raised under regular light cycle but were markedly suppressed if raised in dark (Supplemental Fig. 3A). The total ONL loss during the time course (P16-P60) was in the order as: central retina raised under regular light cycle > peripheral retina raised under regular light cycle > peripheral retina raised in dark > central retina raised in dark, which seemed to correlate with the sum of the FAF area during the time course (Supplemental Fig. 3B). These results suggest that regarding the correlation between the FAF area and the type of photoreceptor apoptosis, the peripheral arcuate FAF may follow the initial event of early photoreceptor apoptosis which occurs regardless of the light condition, while widespread diffuse patterns of apoptosis and increased FAF are suppressed in dark environment, leading to ONL protection.
increased fAf was also observed in retinal degeneration with delayed onset. To investigate the effect of light on the progression of retinal degeneration in rd10 mice, we examined the changes in FAF and cell apoptosis before and after changing the light environment. The rd10 mice were kept in dark from birth to postnatal 10 months and then were moved to an environment with regular light cycle. Significant increase in ssDNA + apoptotic cells was observed 7 days after moving to regular light cycle, which accompanied increase in FAF by in vivo imaging (Supplemental Fig. 4).
A microglial wave spatiotemporally followed the initial apoptotic wave from center to periphery in early retinal degeneration. Arcuate autofluorescent spots and arcuate aggregation of apoptotic cells were observed in the peripheral retina of rd10 mice at P18 under either regular light cycle or continuous dark (Figs. 3A, 5A). Earlier events preceding this phenomenon were sought by examining the spatiotemporal distributions of apoptotic cells and microglia in flat-mounted rd10 retinas from P14 to P19. An initial wave of ssDNA + apoptotic cells appeared on P15 and spread from the central towards the periphery. The wave of Iba1 + microglia appeared on P16-17 in the mid-peripheral retina and then moved towards the periphery following the wave of apoptosis (Fig. 6A). The spreading of apoptotic and microglial waves was significantly faster on the ventral side than the dorsal side (Fig. 6B). In P18 flat-mounted retina, some of the Iba1 + microglia were phagocytizing cells with autofluorescence at 488 nm excitation wavelength, whose localization corresponded to the peripheral arcuate FAF observed in Figs. 3 and 5 (Fig. 6C). In P18 retinal section, the peripheral apoptotic area showed many ssDNA + apoptotic cells exclusively in the ONL (Fig. 6D, E). The area where microglia presented showed a lot of Iba1 + microglia infiltrating into the ONL and reaching to the OS where they showed autofluorescence (Fig. 6D, F). These results suggested that the peripheral arcuate FAF may indicate the event of initial apoptotic wave in the early phase of retinal degeneration in rd10 mice, which occurred regardless of the light condition.

Infiltration of phagocytic microglia into the layers of photoreceptors (ONL and OS).
In the retina of rd10, microglia exhibited two forms: ramified type and amoeboid type. Ramified type microglia were negative for a phagocytic marker CD68. Amoeboid type microglia were positive for CD68 (Supplemental Fig. 5A). The ramified form of microglia resided in the inner and outer plexiform layers (IPL and OPL) in normal retina. In the rd10 retina, microglia did not show any change at P16. From P18 to P22, some microglia became CD68 + amoeboid type, infiltrated toward the ONL, and passed through the ONL to reach the OS. From P30 to P60, the expression of CD68 in microglia subsided (Supplemental Fig. 5B).
Accumulated microglia in the subretinal space are the candidate source of diffuse FAF spots. Since widespread diffuse hyperfluorescent spots by FAF were most evident around P22 under normal light condition, histological examination was performed on rd10 retina isolated at P21. In the flat-mounted central retina, the overall autofluorescence pattern was similar to the accumulation pattern of the Iba1 + cells (Fig. 7A, B). The immunohistochemistry (IHC) data showed that 488 nm autofluorescence colocalized with Iba1 + microglia but not with ssDNA + apoptotic cells (Fig. 7C). In the fundus of rd10 mice on P21, diffuse white spots on the retina were seen by color fundus photograph (Fig. 7D), and these spots emitted autofluorescence at 488 nm excitation wavelength (Fig. 7E). In the retinal section, a number of Iba1 + microglia with autofluorescence were observed in the OS layer (Fig. 7F, F'). In the ONL, amoeboid microglia with or without phagocytosed ssDNA + and ssDNA − photoreceptors were observed, but these microglia in the ONL did not present autofluorescence (Fig. 7G). On the other hand, microglia that reached the OS contained granules with 488 nm autofluorescence in their cell body (Fig. 7H, the substance of the 488 nm autofluorescence will be discussed later). Our results suggest that these microglia accumulated in the OS layer may contribute to the diffuse FAF spots.
Apoptotic photoreceptors express the chemokine, fractalkine. To investigate whether major chemokines are involved in the chemotaxis of microglia in this RP model, qRT-PCR for fractalkine/CX3CL1 and MCP-1/CCL2 was performed on the extracted retina of rd10 mice at 3 different time points during retinal degeneration (P17, 21, and 30). The expression of fractalkine mRNA was lower in rd10 retinas compared to WT except for P21 that showed the expression comparable to WT (Fig. 8A). MCP-1 mRNA was significantly higher in rd10 retina than in WT at P21 and P30 (Supplemental Fig. 6A). www.nature.com/scientificreports/ By IHC, fractalkine + cells were found in the ONL as well as in the RPE layer of rd10 retina at P21, while it was not detected either in the ONL or the RPE layer in WT (Fig. 8B). The staining of fractalkine colocalized with www.nature.com/scientificreports/ some of the ssDNA + apoptotic photoreceptors in the ONL of rd10 retinas (Fig. 8C). The staining of fractalkine colocalized with the area positive for a photoreceptor marker recoverin in the ONL of rd10 retinas (Fig. 8D). These may suggest localized expression of fractalkine in some apoptotic photoreceptors, although the total gene expression level showed no difference compared to WT. By IHC, MCP-1 expression was not detected in WT or rd10 retina at P17 and P21. However, on P30, MCP-1 positive cells were observed in the outer retinal layer (Supplemental Fig. 6B), and some of the Iba1 + microglia were labeled with MCP-1 at the outer retina (Supplemental Fig. 6C), indicating some of the active microglia that migrated to the outer retina was MCP-1 positive.

Contribution of RPE cells to fundus autofluorescence.
It has been assumed that RPE cells, another type of retinal phagocyte, are involved in the FAF in RP patients 3 . Therefore, we examined the autofluorescence of RPE cells histologically. RPE is a monolayer of hexagonal black pigment cells. RPE cells in rd10 mice were apparently similar to those in WT at P18, but on P22 the central RPE cells of rd10 presented dim autofluorescence behind the brighter microglial autofluorescent spots, and showed abnormal morphologies including large size and irregular shapes (Supplemental Fig. 7A, B middle). On P30 they returned to a hexagonal shape, but some RPE cells contained autofluorescent granules in their cell bodies (Supplemental Fig. 7A, B right), which www.nature.com/scientificreports/ looked similar to those in microglia that contained the autofluorescent granules in CD68 + phagosomes (Supplemental Fig. 7C). In the peripheral, although microglia showed autofluorescence as in the central, RPE cells kept the hexagonal shape and did not show autofluorescence (Supplemental Fig. 7A). Altogether, the RPE cells in the central but not the peripheral retina of rd10 were active like microglia, changing their morphologies and accumulating autofluorescent granules in their cell bodies. These results suggest that RPE cells together with microglia contribute to the formation of FAF in RP patients.

Discussion
Whitening and autofluorescent spots were observed on the fundus of rd10 mice. The extent of retinal whitening linearly correlated with the number of apoptotic photoreceptors. Although the time points of the appearance of the autofluorescent spots were different between the central and the peripheral retinas, or whether raised under regular light cycle or in dark, the peak period of FAF appearance overlapped with the peak period of retinal ONL thinning. We therefore hypothesized that microglia that received signals from apoptotic photoreceptors in degenerative retina of RP could migrate into the ONL and contribute to the development of FAF, thus be a biomarker to indicate the progression of RP. In the present study, we obtained the findings that support the hypothesis as follows: (1) Histopathological examination revealed that microglia that phagocytosed apoptotic cells and migrated into the OS layer were the source of autofluorescent spots, (2) From the fundus whitening and autofluorescence intensity, we were able to know the severity of photoreceptor apoptosis and the thinning rate of the ONL, in other words the severity of the ongoing retinal degeneration, and (3) In the retinas of rd10 mice, fractalkine was expressed in apoptotic photoreceptors unlike those of WT. We speculate that fractalkine from apoptotic photoreceptors may induce microglial infiltration into the ONL in rd10 mice. The retina of rd10 on P20 became turbid white. On the next day it turned into white spots with autofluorescence, and those spots were microglia. This suggests microglia are also involved in fundus whitening. The retina is whitened by commotio retinae due to ocular contusion. According to the report of histopathological examination on commotio retinae, a rupture in the OS has been observed 14 . In the retina of rd10 on P20, a large number of microglia infiltrated and reached the OS. Considering this observation together with the report on commotio retinae, we speculate that scattering of reflected light by microglia arriving at the OS, or phagocytosis of the OS by microglia, are related to retinal whitening. Apoptosis of photoreceptors and infiltration of microglia into the ONL has been known to correlate 10 , which is consistent with the results of this study showing that microglia may be involved in fundus whitening and that fundus whitening is correlated with the amount of photoreceptor apoptosis.
In the present study, we observed two patterns of cell apoptosis associated with respective patterns of FAF. Arcuate FAF was formed by early microglial wave following the initial wave of photoreceptor apoptosis spreading from the center to the periphery at the onset of retinal degeneration in rd10 mice (P15-P18). Some microglia that reached the OS emitted autofluorescence in the peripheral retina on P18, which was observed in rd10 mice raised either under regular light cycle or in dark. On the other hand, widespread diffuse FAF peaked around P22 following widespread diffuse apoptosis that peaked around P19-P20. This diffuse pattern was markedly suppressed in rd10 mice raised in dark. Microglia that phagocytized diffusely distributed dying photoreceptors and then reached the OS contributed to the diffuse hyperfluorescent spots. In the retina of RP model mice, the number of microglia infiltrating into the ONL correlates with the number of apoptotic photoreceptors 10 . Together with our results, this suggests that the appearance of FAF spots reflects how much apoptosis has occurred. The short wave FAF signal is known to originate from bisretinoid lipofuscin that is formed in photoreceptor cells as a product of visual cycle reactivity 15,16 . We found that microglia reaching the OS layer became autofluorescent. The emergence of phagocytic microglia/macrophage producing autofluorescence in diseased brain 17 , spinal cord 18 , and retina 6 have been reported. The origin of the autofluorescence of microglia/macrophage has been reported to be redox cofactors such as flavin adenine dinucleotide 19 or lipofuscin granules 20 . Fluorescent materials in the photoreceptors of the OS layer, which were phagocytosed by microglia, were considered to be the origin of the FAF spots. Taken together, we newly showed that microglia along with RPE cells contributed to the formation of FAF that reflect the type and the extent of photoreceptor apoptosis in RP.
In rd10 mice, retinal degeneration seemed to occur with initial wave of apoptosis followed by widespread diffuse apoptosis. In the early stage of retinal degeneration (P15-P19), the wave of photoreceptor apoptosis spread from the center to the periphery. Since this initial apoptotic wave appeared also in the group raised in dark, the light environment might not affect the initial apoptotic wave. On the other hand, widespread diffuse apoptosis was suppressed if raised in dark, and under this condition thinning of the peripheral ONL was severe but the central part was mostly preserved, while under regular light cycle, widespread diffuse apoptosis occurred and degeneration of the central ONL was severe. This indicates the difference of characteristics between the two types of apoptosis: the initial apoptotic wave is independent of the light condition and is involved exclusively in the degeneration of the peripheral ONL, while the later diffuse apoptosis depends on the light condition and is involved in the degeneration of the central ONL. Visual signal triggered by rhodopsin activation is transduced via PDE6 that hydrolyzes cGMP in the rod photoreceptors. Mutation of PDE6β, one of the catalytic units in PDE6, in rd10 mice results in elevated cGMP in the rod photoreceptors 21 , opened cGMP-gated channels and increased intracellular calcium, which leads to apoptosis 22 . It could be possible that photo damage to fragile PDE6β-mutated photoreceptors accelerated the widespread diffuse apoptosis as shown in this study. Similar to the rd10 mice, shading may be effective for retinal protection in some types of RP patients who have mutation in PDE6β.
During the course of degeneration in the retina of rd10, the peak of ONL thinning overlapped with the peak of FAF appearance. Total FAF area over time correlated with the total thinning of the ONL in each group. Genetic ablation of microglia and inhibition of microglial phagocytosis ameliorate photoreceptor degeneration in rd10 mice 10 . This together with the results of our study suggest that recruitment of microglia triggered by apoptosis may further accelerate the ONL loss. Since microglia that infiltrated through the ONL became fluorescent when they reached the OS, the progress of photoreceptor phagocytosis can be estimated by FAF intensity. It is known that hyperautofluorescent ring appears in the fundus of human RP patients. As the retinal degeneration progresses, the hyperautofluorescent ring narrows 3,23 . The hyperautofluorescence corresponds to the boundary of degeneration as revealed by OCT imaging 2 . These together with the results of our study suggest FAF should represent the ongoing ONL loss. www.nature.com/scientificreports/ Although the mRNA expression of fractalkine in the retina of rd10 was not increased compared to WT, immunohistological examination suggested the expression of fractalkine by some apoptotic photoreceptors. In the retina of rd10, microglia expressed CD68, a lysosomal marker indicative of phagocytic activity of microglia 24 , and infiltrated into the layers of photoreceptors (ONL and OS). Microglia expresses the receptor of fractalkine, CX3CR1 25 . In normal retina, fractalkine is expressed in the ganglion cell layer (GCL) and the INL, which coincide with the normal existence site of microglia 26 . Fractalkine is suggested to acts as a "find-me signal", recruiting microglia toward apoptotic cells to promote clearance of the apoptotic cells by phagocytic microglia 27 . In a retinal light-damage model, fractalkine is expressed in apoptotic photoreceptors and attract microglia to the injured ONL 11 . In rd10 mice, fractalkine signal negatively regulates phagocytosis of microglia 28 . These reports together with the results of our study suggest that overall decrease of fractalkine expression in the retina of rd10 activates microglia and altered fractalkine distribution directs the microglia to apoptotic photoreceptors to clear them.
On the other hand, as shown in this study, upregulation of MCP-1 had a time lag from the peak of apoptosis or the appearance of FAF spots. Histologically, MCP-1 was expressed in some microglia that migrated into the outer retina. Activated microglia are able to produce MCP-1 5,12 . The increased expression of MCP-1 in the retina of rd10 may be the consequence of microglia activation after retinal degeneration, leading to neuroinflammation.
We admit there are several limitations in this study. In rd10 mice, the progression speed of retinal degeneration is faster than human RP and the appearance pattern of FAF is different from human RP patients. In this study, we showed the arcuate FAF that appears during the early stage of the degeneration in rd10 mice resembles the hyperfluorescent ring in the fundus of human RP patients in terms of the fact that both demarcates the degeneration area of the photoreceptors. We also showed the involvement of microglia in the FAF formation of rd10 mice. However, it is unclear from the results of this study whether microglia contribute to the hyperfluorescent ring in human RP patients. Successful imaging of microglia in the human retina by adaptive optics imaging has been reported 29 . In the future, it may be possible to get direct evidence with adaptive optics whether microglia contribute to the FAF of human RP patients.
The reason of the difference in the phenotype of FAF between rd10 mice and human RP patients may be the differences in spacial distribution of the rod/cone 30 , the presence or absence of macular structure, the speed of degeneration, and the size of the eye. However, an observation in rd10 mice is significant to interpret human fundus imaging of RP patients because the functional interactions between the retinal cells such as the photoreceptor cells, retinal microglia, and RPE cells should share much in common with those of human.
In conclusion, the fundus whitening and FAF spots were observed in rd10 mice during the course of degeneration. The fundus whitening correlated with apoptosis of the photoreceptors, and the area of FAF spots indicated the thinning rate of the ONL. Thus, fundus whitening and FAF spots can be an indicator for the ongoing severity of degeneration in RP model mice. If the involvement of microglia in the fundus color and autofluorescence is also applicable to human RP, analyzing the fundus color and FAF in detail may help to read the ongoing severity of degeneration in RP patients as well. Histologically, microglia that became phagocytes contributed to FAF formation. The localization of fractalkine to apoptotic cells was suggested to contribute to microglia migration into the ONL. The present results highlight the mechanism and the potential of fundus imaging that might be an indicator of the ongoing disease severity in RP patients.

Methods experimental animals.
All animal experiments were performed according to the guidelines for animal experiments of RIKEN Center for Biosystems Dynamics Research and approved by the Animal Experiment Committee of the RIKEN Kobe Institute (Approval ID: A2006-05-40). All procedures complied with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. WT C57BL/6 J mice were obtained from CLEA Japan, Inc. PDE6β rd10 /J (rd10) mice were obtained from Jackson laboratory, ME, USA and they were housed at a local animal facility under standard laboratory conditions (18-23 °C, 40-65% humidity, and a 12 h light-dark cycle) with free access to food and water throughout the experimental period. A dark-raised group was raised in a dark place for 24 h as a treatment intervention for rd10.
in vivo fundus imaging. After the pupils of mice were dilated with 0.1% Phenylephrin and 0.1% Tropicamid the mice were sedated by inhalation of 1% isoflurane delivered by a nose cone. In vivo shortwave fundus autofluorescent and color images were obtained using the scanning laser ophthalmoscopy; The Micron IV Reti-Scientific RepoRtS | (2020) 10:14700 | https://doi.org/10.1038/s41598-020-71626-2 www.nature.com/scientificreports/ nal Imaging Microscope (Phoenix Research Labs, CA). The field of view as measured from the entrance pupil will be above 50°. The short wave FAF image was obtained by irradiating blue light as excitation light to the retina and separating fluorescence with a 500 nm barrier filter.
imaging analysis. Images were analyzed with imaging processing software ImageJ 32 . (1) Color fundus photography: The area of fundus whitening at each day of age were measured. The 'color threshold' was adjusted to a whitish part, and the %area was calculated using the 'analyze particle' function of ImageJ. (2) FAF imaging: Images were converted into 8-bit images. In raw FAF image, the background is brighter at the center than the periphery and hinders the analysis of the fluorescent spot. For this analysis, the background was excluded with rolling ball function (radius 50 pixels) to leave the fluorescent spot. 'Threshold' was adjusted to the FAF spots and the %area was calculated using the 'analyze particle' function of ImageJ. The area of FAF spots in the central retina at each day of age were measured in the region of 100 × 100 pixels on the dorsal area within the distance of 2 times of papillary diameters from the optic disc. The area of FAF spots in the peripheral retina were measured in the region of 100 × 100 pixels on the dorsal area at the distance of more than 6 times of papillary diameters apart from the optic disc. If arcuate FAF spots appeared in the peripheral retina, the area was measured in the arcuate FAF spots region. (3) Flat-mounted retina: In the analysis of apoptosis, the area of ssDNA positive part of immunostaining in retinal flat mount at each day of age were measured. Images were converted into 8-bit images. 'Threshold' was adjusted to ssDNA + parts and the %area was calculated using the 'analyze particle' function of ImageJ.
Statistical evaluation. Data representing the mean ± standard deviation for the results of at least three independent experiments were compared by Student's t test. Values were considered statistically significant if p was less than 0.05. The regression line between area ratio of ssDNA + in the ONL and area ratio of fundus whitening was calculated. Statistical analysis was conducted using the commercial software package Excel (Microsoft, WA, USA) and Statcel3 (Add-in forms on Excel, H. Yanai, OMS, Tokyo, Japan).

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.