In vivo analysis of onset and progression of retinal degeneration in the Nr2e3rd7/rd7 mouse model of enhanced S-cone sensitivity syndrome

The photoreceptor-specific nuclear receptor Nr2e3 is not expressed in Nr2e3rd7/rd7 mice, a mouse model of the recessively inherited retinal degeneration enhanced S-cone sensitivity syndrome (ESCS). We characterized in detail C57BL/6J Nr2e3rd7/rd7 mice in vivo by fundus photography, optical coherence tomography and fluorescein angiography and, post mortem, by histology and immunohistochemistry. White retinal spots and so-called ‘rosettes’ first appear at postnatal day (P) 12 in the dorsal retina and reach maximal expansion at P21. The highest density in ‘rosettes’ is observed within a region located between 100 and 350 µM from the optic nerve head. ‘Rosettes’ disappear between 9 to 12 months. Non-apoptotic cell death markers are detected during the slow photoreceptor degeneration, at a rate of an approximately 3% reduction of outer nuclear layer thickness per month, as observed from 7 to 31 months of age. In vivo analysis of Nr2e3rd7/rd7 Cx3cr1gfp/+ retinas identified microglial cells within ‘rosettes’ from P21 on. Subretinal macrophages were observed in vivo and by confocal microscopy earliest in 12-months-old Nr2e3rd7/rd7 retinas. At P21, S-opsin expression and the number of S-opsin expressing dorsal cones was increased. The dorso-ventral M-cone gradient was present in Nr2e3rd7/rd7 retinas, but M-opsin expression and M-opsin expressing cones were decreased. Retinal vasculature was normal.


Results
Onset of retinal degeneration in C57BL/6J Nr2e3 rd7/rd7 mice. To assess in vivo the appearance of retinal white spots and 'rosettes' in C57BL/6J Nr2e3 rd7/rd7 retinas during early postnatal development, we performed fundus examination and optical coherence tomography (OCT). White spots appeared first dorsally at postnatal day 12 (P12) in C57BL/6J Nr2e3 rd7/rd7 retinas (Fig. 1A). The number of white spots and 'rosettes' increased between P12 and P13 (Fig. 1B). A centrifugal expansion of white spots was then observed between P15 and P21 ( Fig. 1C-E). Detailed fundus examination and OCT at P28 showed a high density of intense white spots from central to peripheral regions with peripapillary sparing (Fig. 1F), and a gradual decrease in white spots and 'rosettes' towards the far periphery (Fig. 1G). At this time-point we could also accurately determine that all white spots on fundus examination exactly correlated with the presence of 'rosettes' in the outer nuclear layer (ONL) (Fig. 1H). No 'rosettes' were detected at any time-point in C57BL/6J retinas (Fig. S1).
To further describe and quantify the early retinal phenotypes of C57BL/6J Nr2e3 rd7/rd7 mice we resorted to histology (Fig. S2). At P12, the outer retina appeared wavy and 'rosettes' were not fully developed (Fig. S2B). At P13, 'rosettes' were formed and protruded towards the inner nuclear layer (INL) (Fig. S2D). The number of 'rosettes' increased dramatically between P12 and P13: per histological section passing through the optic nerve head, 1.2 'rosettes' were observed on average at P12, and 12.2 at P13 (Fig. 1I). The average number of 'rosettes' increased to 14.3 at P18, 17.5 at P21 and 18.7 at P28. At P35, the average number of 'rosettes' had decreased to 12.5 on histological sections. We also performed a qualitative analysis of 'rosette' localization along a dorsoventral axis passing through the optic nerve head (Fig. 1J). After appearance of 'rosettes' at P12 on the dorsal side, a pan-retinal expansion of 'rosettes' was already present at P13. At P18, 'rosettes' had expanded towards the periphery and reached maximal expansion at P21, with 'rosettes' observed at a distance of more than 1.5 mm from the optic disk. At all time-points and independently from the expansion towards the periphery, the highest density of 'rosettes' was observed in a near-central region, within a distance of 100-350 µM from the optic disk head. Typically, no 'rosettes' were observed in the immediate vicinity of the optic disk and in the far periphery. No 'rosettes' were observed at P0, P3, P5, P8 and P11 in C57BL/6J Nr2e3 rd7/rd7 mice, and in any C57BL/6J and heterozygous C57BL/6J Nr2e3 rd7/+ mice analyzed at these time-points (data not shown).
To analyze retinal microglia in vivo, we crossed C57BL/6J Nr2e3 rd7/rd7 mice with heterozygous Balb/c Cx3cr1 gfp/+ 'knock-in' mice, that express selectively the green fluorescent protein (GFP) in microglial cells under the control of the endogenous Cx3cr1 locus. We observed a decreased overall number of 'rosettes' in this Nr2e3 rd7/rd7 Cx3cr1 gfp/+ mouse line of mixed genetic background in comparison to the C57BL/6J background (data not shown). Fundus fluorescence imaging readily detected GFP + retinal microglial cells (Fig. 3A). By OCT we could colocalize GFP + microglial cells and 'rosettes' and detect there both an increased fluorescent signal and an increased density of GFP + microglial cells (Fig. 3B,C). These spots of increased GFP fluorescence were observed in Nr2e3 rd7/rd7 Cx3cr1 gfp/+ retinas (Fig. 3D), but not 'wild-type' Cx3cr1 gfp/+ ones (Fig. 3E). Longitudinal analysis of GFP + microglial cells in a same retina allowed to detect GFP + microglia at 'rosettes' from P21 to P60, both in persisting and newly formed 'rosettes' (Fig. 3F-H).
Mild progression of retinal degeneration in old C57BL/6J Nr2e3 rd7/rd7 mice. Fundus photography in 7-and 9-month-old mice showed that a few hyperreflective white spots were still present in the central to midperipheral retina, colocalizing with 'rosettes' expanding to the INL on OCT (Fig. 4A,B). These bright white spots on fundus had disappeared at 12 months of age, as well as 'rosettes' on OCT (Fig. 4C). However, we observed by fundus photography densely scattered small beige-yellow spots, that were already visible at 7 months of age and persisted to later time-points ( Fig. 4A-D). These small spots were not observed in C57BL/6J retinas (Fig. 4E,F  Per retina, the number of 'rosettes' was counted on five hematoxylin-eosin stained sections along a dorsoventral axis containing the optic nerve head. Statistical analysis was performed by ordinary one-way ANOVA with Tukey's multiple comparisons test. *p < 0.05 ***p < 0.001; ****p < 0.0001. (J) Qualitative spatio-temporal distribution of rosettes analyzed in graph I along a dorso-ventral axis relative to the optic nerve head (0). www.nature.com/scientificreports/ On these aging mice, we performed a longitudinal analysis of the ONL thickness by OCT (Fig. 4G). As a reference we used the ONL thickness of C57BL/6J mice at 6 (N = 38) and 18 months (N = 29), where a nonsignificant decrease by 3.95% from 62.96 ± 0.453 µM to 60.47 ± 0.4902 µM was measured. In comparison to 6-month-old wild-type mice, the ONL thickness in C57BL/6J Nr2e3 rd7/rd7 retinas was decreased by 5.7% at 7 months of age (59.37 ± 0.3917 µM; N = 78). By 15 months of age ONL thickness was further reduced by 21.7% (46.47 ± 0.4362 µM; N = 83). Based on this 8-month period of monitoring, we could estimate the rate of ONL thinning to less than 3% per month. Of note, the inner segment/outer segment boundary was detected in C57BL/6J Nr2e3 rd7/rd7 retinas on OCT at all time-points ( Fig. 4A-D), indicating functional photoreceptors.
S-and M-opsin expression in C57BL/6J Nr2e3 rd7/rd7 retinas. Because of the impaired photoreceptor development in Nr2e3 rd7/rd7 retinas, we assessed in more detail S-and M-opsin expression during early postnatal development. At P12, S-opsin was detected in C57BL/6J Nr2e3 rd7/rd7 retinas at waves starting to invaginate (Fig. 7A), and then in the fully formed 'rosettes' at P13 (Fig. 7B) and P21 (Fig. 7C). M-opsin was not detected in 'rosettes' at P13 (Fig. 7F), but at P21 all 'rosettes' along the dorso-ventral gradient expressed M-opsin (Fig. 7G,K). We then performed Western blot analysis on extracts from P21 retinas (Fig. 7L). Consistent with reported data at other time-points 8,9 , S-opsin expression was increased by 2.6-fold in C57BL/6J Nr2e3 rd7/rd7 retinas in comparison to C57BL/6J levels, whereas M-opsin expression was decreased by 45% (Fig. 7M). As assessed by cone outer segment staining for S-and M-opsin on flat mounts of P21 retinas, the increase in S-opsin pro-    www.nature.com/scientificreports/ tein expression correlated with a 4.8-fold increase in S-opsin expressing cone outer segments in dorsal regions of C57BL/6J Nr2e3 rd7/rd7 retinas, but not in ventral ones (Fig. 7N). We also observed an increased number of dorsally located cones expressing both S-and M-opsin. Conversely, a decrease in M-Opsin expressing outer segments was observed both in dorsal and ventral regions, by respectively 38% and 63%.

Discussion
The detailed analysis of C57BL/6J Nr2e3 rd7/rd7 retinas identified correlations between the topographic distribution of 'rosettes' , the maturation of photoreceptor outer segments and the density of rod photoreceptors. We detected white spots and 'rosettes' first at P12 on the dorsal side, then there was a centrifugal progression within a day over the entire retina reaching the most peripheral expansion at P21 (Fig. 8A). This time frame fits the central to peripheral gradient of photoreceptor maturation, where murine rod outer segments elongate at a rapid and almost linear rate from P11 to P17, and reach adult length by P19-P25 [20][21][22] . Murine rods are very small and packed at a high overall average density of 437,000/mm 223 . Quantitative analysis of photoreceptor distribution in adult C57BL/6 retinas along a 4.8 mm dorso-ventral diameter had shown an about 1.2-fold increase in photoreceptor density from the posterior pole towards a region at about 600 µM from the posterior pole, and then an about 1.4-fold decrease towards the periphery 23,24 . We observed the highest density of 'rosettes' within a region located within 100-350 µM from the optic nerve head along a 3.6 mm dorso-ventral diameter in young postnatal eyes, that corresponds to this region of higher photoreceptor density. In C57BL/6J Nr2e3 rd7/rd7 retinas the dense packing of photoreceptors is exacerbated by an increase in 'cod' cell body size by up to 30% 10 . Additionally, the observed 4.8-fold increase in dorsal cones expressing S-opsin may further exacerbate spatial constraints and the first appearance of 'rosettes' in the dorsal retina at P12 related to S-cone outer segment development. Additional dorsal S-cones and photoreceptor outer segment maturation are plausible causes of the initiation of 'rosette' formation and expansion towards the periphery. In presence of high photoreceptor density and larger photoreceptor cell bodies spatial constraints are then further exacerbated. 'Rosette' formation can be regarded as an attempt to accommodate larger photoreceptor cell bodies for a same RPE surface by 'pushing' photoreceptors towards the periphery. In support of this spatial constraint hypothesis, the genetic removal of cones was sufficient to generate enough space in the retina and prevent 'rosette' formation in Nr2e3 rd7/rd7 retinas 17 . Consistent with 'rosette' formation driven by rod density, patients affected by recessive NR2E3-linked ESCS typically exhibit pathological fundus and autofluorescence changes in the perimacular-to-mid-peripheral region [25][26][27] , where rods are at their highest density of 160,000/mm 228 . Whether this mid-peripheral region also contains an increased number of S-cones in ESCS patients remains elusive 29 .
With respect to cone opsin expression, we postulate that the respective increase in S-opsin and decrease in M-opsin expression we assessed in Nr2e3 rd7/rd7 retinas, reflects the reported ERG findings of enhanced S-cone function but decreased M-cone function in human patients 30,31 . www.nature.com/scientificreports/ White spots and 'rosettes' gradually disappeared between 9 and 12 months of age in adult C57BL/6J Nr2e3 rd7/rd7 retinas. We also determined an approximately 3% decrease per month in ONL thickness in the C57BL/6J Nr2e3 rd7/rd7 retina between 7 and 31 months of age. This is consistent with retinal function within normal limit until 5 months of age, but reduced by 50% at 16 months as assessed by electroretinography (ERG) 16 . The slow-progressing retinal degeneration observed in C57BL/6J Nr2e3 rd7/rd7 mice resembles to what is observed in ESCS patients 32 . Because we detected increased expression of the non-apoptotic cell death markers PAR, Calpain-2 and Survivin, but not of the pro-apoptotic Bax, cleaved Caspase 3 and cleaved PARP-1 and the pronecroptotic MLKL markers, our data suggest that the slow retinal degeneration observed in aging C57BL/6J Nr2e3 rd7/rd7 retinas is driven by non-apoptotic cell death pathways, similar to what is observed in a vast majority of analyzed murine hereditary retinal degeneration models 33 .
The presence of immune cells within 'rosettes' may mediate the waste removal of trapped photoreceptor outer segments, normally phagocytosed by the RPE 18,19 . We observed an increased fluorescence signal in vivo in microglial cells around rosettes, which would be consistent with an increased uptake of hyperautofluorescent photoreceptor outer segments. Importantly, our analyses show an early phase of microglial cell migration into 'rosettes' , followed by the immigration of monocytes/macrophages across the RPE (Fig. 8B). The small yellow spots observed by fundus photography in retinas of old mice have been associated with subretinal microglial cells 34 , but our immunohistochemical analysis detecting subretinal localization of F4/80 + cells in aged C57BL/6J Nr2e3 rd7/rd7 retinas is suggestive of subretinal macrophages. We hypothesize that the subretinal fibrosis described in ESCS may be caused by the presence of subretinal macrophages 35 . Further analyses will be necessary to establish the respective roles of infiltrated macrophages and resident activated microglia in this retinal degeneration.
Taken together, our data identified additional S-cones and photoreceptor outer segment maturation as likely triggers of 'rosette' formation. Initial microglia migration towards 'rosettes' is followed by monocyte/macrophage immigration. These findings further illustrate the validity of the Nr2e3 rd7/rd7 mouse retina to study ESCS-associated disease mechanisms 36 .

Animals.
All experiments performed in this study were in accordance with the ARRIVE guidelines and were approved by the Veterinary Offices of the Canton of Bern (authorization BE17/19). C57BL/6J (RCC, Basel, Switzerland) and B6.Cg-Nr2e3 rd7 /J mice (Jackson Laboratory, Bar Harbor, ME, USA) were kept in a 12-h light-dark cycle with unlimited access to food and water. In order to increase the fertility of the B6.Cg-Nr2e3 rd7 /J mice that had been backcrossed over 8 generations to homozygosity at Jackson Laboratory and had a typical litter size of 1-3 pups, we backcrossed them over 4 generations to the C57BL/6J genetic background, resulting in our C57BL/6J Nr2e3 rd7/rd7 line with an average litter size of 6-8 pups. During backcrossing, all litters were systematically checked for the presence of the rd1 and rd8 mutations 37,38 . C57BL/6J Nr2e3 rd7/rd7 mice were further crossed with homozygous mice selectively expressing green fluorescent protein (GFP) in microglia under the control of the Cx3cr1 gene were obtained by crossbreeding wild type Balb/cAnNCrl females with male transgenic homozygous fractalkine receptor reporter mice (Cx3cr1 gfp/gfp ) on a Balb/c background 39 . In vivo imaging. Mice   Retinal flat mount. Retinas were isolated under a dissecting microscope and gently transferred to a 96-well cell culture dish containing 4% paraformaldehyde-1xPBS for 2 h. They were washed three times for 5 min in 1xPBS and blocked for 1 h at room temperature with gentle agitation in blocking buffer (2% normal goat or horse serum, 0.2% Triton X-100 in 1xPBS). Incubation in primary antibody (OPN1SW, sc-14365, Santa Cruz, Dallas, TX, USA, 1/100 in blocking solution; OPN1MW, AB5405, Merck Millipore, Darmstadt, Germany, 1/1000 in blocking solution) was performed overnight at 4 °C with gentle agitation. Secondary antibodies, diluted 1/1000 in blocking solution, were incubated for 1 h at room temperature. Retinas were then washed 3 times in 1xPBS and transferred to a glass slide. To flatten the retinas, four cuts were made, equidistant apart, from the periphery of the retina towards the center. Slides were mounted in Citifluor before images were acquired on a Leica DM 6000B microscope. Cone outer segments were quantified using Fiji/ImageJ version 1.51 (http:// imagej. nih. gov/ ij; National Institutes of Health, Bethesda, Maryland USA). Measurements were done in dorsal and ventral regions on rectangles of 570 × 1370 pixels. Background was removed using in-built Otsu thresholding method with analyzed particle size (volume) set to 10-400 pixel 2 .

Statistical analysis.
All results were expressed as means ± SD or ± SEM, and the number of samples and experiments indicated in text and figure legends. Statistical analyses were performed with Prism 8.2.0 (Graph-Pad Software Inc., La Jolla, CA, USA).