Sfrp1 deficiency makes retinal photoreceptors prone to degeneration

Millions of individuals worldwide suffer from impaired vision, a condition with multiple origins that often impinge upon the light sensing cells of the retina, the photoreceptors, affecting their integrity. The molecular components contributing to this integrity are however not yet fully understood. Here we have asked whether Secreted Frizzled Related Protein 1 (SFRP1) may be one of such factors. SFRP1 has a context-dependent function as modulator of Wnt signalling or of the proteolytic activity of A Disintegrin And Metalloproteases (ADAM) 10, a main regulator of neural cell-cell communication. We report that in Sfrp1−/− mice, the outer limiting membrane (OLM) is discontinuous and the photoreceptors disorganized and more prone to light-induced damage. Sfrp1 loss significantly enhances the effect of the Rpe65Leu450Leu genetic variant -present in the mouse genetic background- which confers sensitivity to light-induced stress. These alterations worsen with age, affect visual function and are associated to an increased proteolysis of Protocadherin 21 (PCDH21), localized at the photoreceptor outer segment, and N-cadherin, an OLM component. We thus propose that SFRP1 contributes to photoreceptor fitness with a mechanism that involves the maintenance of OLM integrity. These conclusions are discussed in view of the broader implication of SFRP1 in neurodegeneration and aging.


Results
Young adult Sfrp1 −/− retinas show subtle defects in cone photoreceptor organization. The final goal of our study was to determine whether SFRP1 is part of the molecular machinery that maintains photoreceptors' integrity. We reasoned that, if this is the case, mice lacking SFRP1 activity (Sfrp1 −/− ) should have a different susceptibility than their wild type (wt) counterparts to light-induced damage, taken as a frequently used experimental paradigm to study photoreceptor degeneration 29 . Sfrp1 does not seem to be required for mouse retinal development 18,30,31 . However, there are somewhat controversial reports on its adult retinal localization 32,33 and, to our knowledge, no specific information on its possible function in adult retinal homeostasis. We addressed these issues first.
To clarify Sfrp1 distribution, we hybridized sections of 1 month-old mouse eyes with a specific probe. Our results supported the report by Liu et al. 32 . High levels of Sfrp1 were localized to the inner nuclear (INL) and ganglion cell (GCL) layers. Lower levels were also found in the outer nuclear (or photoreceptor) layer (ONL; Fig. 1a), with a more abundant distribution in a subset of cells located at the outermost region (Fig. 1b, white arrows), in which the cell bodies of cone photoreceptors are found 34 . Immunostaining with specific antibodies confirmed a similar layered distribution of the protein that, according to its secreted and dispersible nature 18,35 , was notably accumulated at the OLM/OS region (Fig. 1c,d). ELISA determination of SFRP1 content in retinal extracts from mice of age comprised between 1 and 25 months showed that its levels significantly decreased with age (Fig. 1k).
In situ hybridization analysis did not detect the expression of the closely related Sfrp2 in the adult mouse retina as described for humans 27 , whereas Sfrp5 reporter expression was localized only in few sparse retinal cells and in the retinal pigmented epithelium (Fig. S1).
To determine if the activity of SFRP1 is required for the organization of the adult retina, we compared the retinal morphology of 2 months-old Sfrp1 −/− and wt mice using toluidine blue stained semi-thin sections. No appreciable gross differences were detected between the two genotypes and the thickness of the ONL layer was comparable in both the dorsal and ventral quadrants (Fig. 1e,f), even in 4 months-old mice ( Fig. 1l; n = 3 mice/ genotype, Mann-Whitney U test). In mice, cone cell bodies are mostly arranged in a single row below the OLM with nuclei characterized by the presence of 1-3 irregular clumps of heterochromatin. Rods instead occupy all the ONL rows with more densely stained nuclei and a single large central clump of heterochromatin 34 . This arrangement was easily recognizable in wt retinas observed at higher magnification (Fig. 1g), with 86.1% (±0.2) of the cones localized to the first or second row of the ONL, whereas the remaining 13.9% (±2.9) were found in other rows (n = 3 mice). In Sfrp1 −/− retinas instead, cone nuclei were abnormally distributed (Fig. 1h, red asterisks) and only 51.1% (±6.2) of the cone cell bodies were found in the first two rows. The remaining 48.9% (±3.1) were distributed in other rows, with a significant cone increase in deeper rows (n = 3 mice; Chi-square test, p < 0.001). Nonetheless, the total number of cones was not significantly different between wt (291 ± 8, in nerve head sections, n = 3) and Sfrp1 −/− mice (336 ± 13.5, n = 3; Mann-Whitney U test), at least in young adults. Nevertheless, ultra-structural analysis revealed occasional degenerative signs in the OS of the mutants, which were never observed in wt (Fig. 1i,j).
Sfrp1 −/− mice were bred in a mixed C57BL/6 ×129 background. Previous studies have warned that C57BL/6 N sub-strain carries a single nucleotide deletion in the Crumb homologs 1 (Crb1) gene, which causes a retinal phenotype that may override that of other genes of interest 36 . Our animals were bread in the C57BL/6 J sub-strain, reported to carry a wt allele 36 . Nevertheless, and to confirm that the observed defects were bona fide associated to the loss of Sfrp1 function, we sequenced the Crb1 gene in the Sfrp1 −/− line. No Crb1 mutations were found in all the analysed samples of our colony (not shown).
Once excluded the possible contribution of a defective Crb1 allele, we next asked if the subtle defects observed in Sfrp1 −/− retinas had an impact on photo-transduction or synaptic transmission between photoreceptors and bipolar cells. To this end, we compared the electrical responses of groups of 4-months old wt and Sfrp1 −/− mice (n = 12 per genotype) to light stimuli of increasing intensity and under increasing frequency. ERGs were recorded under scotopic (Fig. 2a-d) and photopic ( Fig. 2e-g) conditions to evaluate rod/rod bipolar cells and cone/cone bipolar function, respectively. Mutant mice showed a small but statistically non-significant (Mann-Whitney U test) decrease of photoreceptor (a-wave, see Fig. S2) scotopic (Fig. 2c,d) and photopic (Fig. 1f) response at higher intensities and a tendency towards a lower response under flicker stimuli of low frequency (Fig. 2g). (e-h) Semi-thin frontal sections from 2 months-old wt and Sfrp1 −/− retinas stained with toluidine blue. Note the abnormal localization of cone nuclei in the mutant retina (red asterisks in h). (i,j) TEM analysis of the OS. Note that in the mutants, but not in wt, the stacks of membranes of the OS are disorganized (area enclosed in dotted line in (j). (k) ELISA analysis of SFRP1 levels in total protein extracts from 1 to 25 months-old retinas from wt mice. Note the significant decrease of SFRP1 content as the animals age; Oneway ANOVA, with post hoc Bonferroni analysis; *p < 0.05, **p < 0.01. (l) The graph shows the ONL thickness in 4 months-old wt and Sfrp1 −/− retinas (measures were taken in cryostat frontal sections of the eye at the level of the optic disk). No significant differences were detected between wt and Sfrp1 −/− retinas (Mann-Withney U test). Abbreviations: GCL: ganglion cell layer; INL: inner nuclear layer; on: optic nerve; ONL: outer nuclear layer.
www.nature.com/scientificreports www.nature.com/scientificreports/ Together these data suggest that SFRP1 contributes to maintain retinal organization but its absence has negligible effects on young adults.
Retinal disorganization and visual performance of Sfrp1 −/− retinas worsen with age. Photoreceptors disorganization, decreased OS length and scotopic responses together with reduced synaptic density and loss of melanin granules are among the alterations found in the aging mammalian retinas [37][38][39] . The retinal phenotype of the young adult Sfrp1 −/− mice shared some of these features, suggesting that the observed features could worsen with time.
To address this possibility gross retinal morphology of 6 months-old wt and Sfrp1 −/− mice was analysed, finding no considerable macroscopic differences. However, the well-aligned columnar organization of the photoreceptors observed in wt was lost in the Sfrp1 −/− retinas and, cone photoreceptor nuclei were distributed throughout the width of the ONL (Fig. 3a,b), as we observed in young mice. Many photoreceptors' nuclei were also misplaced past the OLM that normally overlays the ONL (n = 6, Fig. 3c,d). Retinal immunostaining with the Transducin/GNB3 and rhodopsin (Rho), markers of the OS of cones and rods, respectively, confirmed subtle photoreceptors' alterations. Transducin/GNB3 staining was particularly reduced and fragmented in the mutant retinas (n = 3 mice for each age, Fig. 3e,f), reflecting a small but significant reduction of the number of cone nuclei counted in toluidine blue stained sections (wt, 231.6 ± 12 vs. Sfrp1 −/− 149 ± 26.7, in retinal nerve head sections; n = 3 and 5 mice, Mann Whitney U test, p = 0.025). In wt mice, Rho normally localises to the OS (Fig. 3g), but in the Sfrp1 −/− retinas labelling was also observed in the cell bodies of a number of cells (Fig. 3h). Although ONL cell density was an obstacle for precise evaluation of the proportion of affected rods, cells with abnormal Rho distribution were consistently observed in both eyes of all the analysed animals (n = 3 mice for each age analysed, i.e., 6 and 8 month).
The presence of Rho in the cell bodies reflects cell intrinsic or OLM disturbances 40 . Indeed, the OLM seals off the photoreceptor inner and outer segments from the rest of the cell, thereby limiting the diffusion of the photo-transduction cascade components 41 . This notion together with the displacement of the photoreceptor nuclei (Fig. 3c,d), similar to that observed upon pharmacological disruption of the OLM 42,43 , suggested the existence of OLM alterations in the mutants. The intermittent breakage of the OLM already apparent in histological sections (red arrows in Fig. 3d) was confirmed by ultrastructural analysis (n = 3). The well-organized arrangement of zonula adherens junctional complexes between the plasma membranes of the photoreceptor ISs and the apical processes of Müller glia that compose the OLM was readily visible in wt retinas (Fig. 4a). This arrangement was no longer present in those regions of the Sfrp1 −/− retinas in which photoreceptors' nuclei were displaced bearing signs of degeneration in both the cell body ( Fig. 4b) and OS. Defective outer segments were more abundant than those observed at 2 months of age (Fig. 1j) and mostly localized in the central retina as determined by quantification in semi-thin sections (Fig. 4c,d; 5 ± 1 in wt and 25 ± 3 in Sfrp1 −/− nerve head sections, n = 3, Mann-Whitney U test, p = 0.043). In Sfrp1 −/− retinas, RPE cells presented less and smaller melanosomes (Fig. 4e,f) and synaptic contacts between photoreceptors and bipolar interneurons were also altered, with an abnormal accumulation of synaptic vesicles (Fig. 4h) and the degeneration of pre and post-synaptic terminals (red arrows in Fig. 4i,j). None of these defects were observed in wt retinas ( Fig. 4g) but have been reported in aging retinas [37][38][39] .
Recording of ERGs responses (Fig. S2) under scotopic (Fig. 4k,l) and photopic (Fig. 4m,n) conditions in 9 months-old wt and Sfrp1 −/− mice showed that the above described morphological defects of Sfrp1 −/− mice were associated with a poorer performance of rods as compared to wt mice ( Fig. 4k; n = 5 per genotype, Mann-Whitney U test, p = 0.035). Direct comparison of the photopic responses of 4 and 9 months-old Sfrp1 −/− mice indicated a progressive deterioration with aging ( Fig. 4n; Mann-Whitney U test, p = 0.033).
Morphological analysis of 12 months-old mice showed worsening of photoreceptor damage with enlarged and degenerated OS and an evident breakage of the OLM (Fig. 4o). Overall, the thickness of ONL was significantly reduced (20% on average) across the almost entire dorso-ventral axis of the retinas as compared to age-matched wt ( Fig. 4p; n = 5 per genotype, Mann-Whitney U test, p < 0.05).
Together these data indicate that Sfrp1 loss causes a slow but progressive deterioration of the retinal integrity associated with decrease of visual function.

The proteolysis of OLM and OS components is increased in Sfrp1 −/− retinas. The region of the
OLM is enriched in cell adhesion molecules, such as N-cadherin found at the adherens junctions 44 and protocadherin PCDH21, located at the base of the OS 45 . Other transmembrane molecules are, for example, CRB2 and CRB1, found at the subapical region (right above the adherens junctions) of the photoreceptors and/or Müller glial cells as part of a large protein complex 6,46 . As already mentioned, a tight adhesion between IS of the photoreceptors and the end feet of the Müller cells is critical for photoreceptor integrity and retinal organization 47 . Furthermore, variation of the human CRB1 and CRB2 genes are responsible of retinal dystrophies such as RP 48 and inactivation of their murine homologues disrupts the OLM and causes photoreceptor loss 49 . We thus reasoned that the retinal phenotype of the adult Sfrp1 −/− mice could be mechanistically linked to poor Müller/ photoreceptor cell adhesion, which, in turn, would be responsible of photoreceptor alterations. This mechanism is based on the notion that SFRP1 acts as a negative modulator of ADAM10 12,13,22,25 , which proteolytically removes the ectodomains (ectodomain shedding) of a large number membrane proteins 8 , including cell adhesion molecules such as Cadherins and proto-Cadherins 14,50,51 , thereby controlling cell-cell interactions 8,51 . Consistent with this idea, it has been shown that abnormal proteolytical processing of OS components influences photoreceptor degeneration 52 . Furthermore, different ADAM proteins, including ADAM10, have been found expressed in different retinal layers and cell types, comprising Müller cells 53,54 . We confirmed this distribution with immunohistochemical analysis of ADAM10 localization. In 3 months-old retinas ADAM10 was specifically found in ganglion cells, in the INL (presumably in Müller glial cells based on position) and in the OLM (Fig. 5a,b). Thus and given that SFRP1 is particularly abundant in the OLM/OS region (Fig. 1c,d), we hypothesized that, in SFRP1 absence, components of the OLM could be proteolyzed more efficiently.
We therefore tested if an enhanced proteolytical processing of N-cadherin and/or PCDH21 could explain the retinal alterations of Sfrp1 −/− mice. To this end, we analysed using Western blot the proportion of proteolyzed versus full-length protein levels of N-cadherin and PCDH21 in wt, Sfrp1 −/− and rd10 mice. These mice, carrying a missense mutation in the rod specific Pde6b, are a widely used model of fast photoreceptor degeneration 55 and thus appeared useful to assess mechanistic specificity. Proteolytical processing of full-length N-cadherin (135 kD) results in the generation of a C-terminal intracellular peptide of 35 kDa 51 , whereas the 120 kDa full length PCDH21 is cleaved in two fragments of 95 (N-terminal) and 25 (C-terminal) kDa 52 . Antibodies against the C-terminal portion of either N-Cadherin or PCDH21 were probed against retinal extracts of 3 weeks-old mice. At this stage, the majority of photoreceptors in rd10 mice are still alive, allowing for comparison. Sfrp1 −/− retinal extracts contained an increased proportion of the C-terminal fragments of both N-cadherin and PCDH21 when compared to those detected in wt (Figs. 5c,d; S3, about 2-and 3.5-fold, respectively). N-cadherin but not PCDH21 proteolysis was also observed in rd10 retinas (Figs. 5c,d; S3). Comparable and unprocessed levels of CRB1 were present in the 3 genotypes (Figs. 5e; S3), supporting processing specificity. Furthermore, the signal of immunostaining for N-cadherin and β-catenin -which functionally interacts with cadherins at the adherens junctions 56 -appeared discontinuous and spread along the Müller glial processes in the Sfrp1 −/− retinas (Fig. 5g,i) but as a continuous line in wt retinas (Fig. 5f,h). Staining of Müller glial cells with anti-Glutamine Synthetase (GS) confirmed the discontinuous arrangement of their apical end-feet in the mutants but not in the wt retinas (Fig. 5j,k).
Altogether these data indicate that SFRP1 indirectly regulates photoreceptors/Müller glia interaction, which is, in turn, essential for retinal integrity and optimal function. Sfrp1 bestows resistance to the photoreceptors upon light induced damage. If SFRP1 maintains photoreceptor integrity, in its absence photoreceptors should be more sensitive to a neurodegenerative stimulus. Exposure to high intensity light is an experimental paradigm widely used to induce a rhodopsin mediated www.nature.com/scientificreports www.nature.com/scientificreports/ stereotypic photoreceptor apoptosis 29 , to which cones are less susceptible 57,58 . The degree of response however varies according to the pigmentation and genetic background of the animals 29 . Most prominently, a Leu450Met variation in the RPE-specific Rpe65 gene, found in the C57BL/6 and 129 mouse strains 59 , confers resistance to light damage. Given the genetic background of the Sfrp1 −/− mice, we sequenced the Rpe65 gene in our colony, finding the presence the Rpe65 Leu450Met variant mostly in heterozygosis. To clearly separate the contribution of the two alleles we generated Rpe65 Leu450Leu ;Sfrp1 −/− and Rpe65e Leu450Met ;Sfrp1 −/− animals and their respective controls. We thereafter exposed 4 months-old wt and Sfrp1 −/− mice carrying the two Rpe65 variants to a 15.000 lux light source for 8 h and analysed the extend of photoreceptors' loss by immunohistochemical analysis 7 days after treatment. Under these conditions and compared to untreated animals (the two Rpe65 variant were undistinguishable), the retinas of Rpe65e Leu450Met ;Sfrp +/+ mice were only slightly affected. The morphology and thickness of their ONL as well as the immunolabelling for rod (Rho) and cones (GNB3) markers were comparable to that of untreated animals (Fig. 6a,c,e,g,m). Given the absence of the resistance allele, the photoreceptors of Rpe65e Leu450Leu ;Sfrp +/+ mice instead presented signs of disorganization and the ONL was slightly thinner in central regions as compared to untreated mice (Fig. 6a,c,i,k,n). By contrast, the retinas of Rpe65e Leu450Met ;Sfrp1 −/− and, to a larger extend, those of Rpe65 Leu450Leu ;Sfrp1 −/− mice were visibly affected by light damage. There was a central high to peripheral low loss of both rod and cones photoreceptors with abnormal localisation of Rho in the cell bodies (Fig. 6b,d,f,h,j,l,n). Cone loss appeared proportionally more significant than that of rods in mice carrying the Rpe65 Leu450Leu vs the Rpe65e Leu450Met variant (compare Fig. 6f,h with j,l; Kruskal Wallis with post hoc Dunn, p < 0.05).
Consistently the reduced loss of photoreceptors observed in Rpe65e Leu450Leu ;Sfrp +/+ after damage (Fig. 6i,k), both scotopic and photopic responses were only slightly and non-significantly reduced in these animals (Fig. 7a-d).
All in all, these data indicate that SFRP1 activity critically supports photoreceptor integrity and function upon a neurodegenerative stimulus.

Discussion
Photoreceptors' degeneration has devastating consequences on visual perception. So far lost photoreceptors cannot be effectively replaced but, in many cases, restraining their alterations would be sufficient to enable a normal life in affected individuals 60 . This may be particularly important for cones because they are activated by day-light or artificial illumination and thus enable sufficient vision in current society. The identification of the mechanisms that contributes to the integrity of photoreceptors (cones in particular) may thus help designing protective/therapeutic strategies to guarantee sufficient visual percetion. In this study, we have identified SFRP1 as a component of the machinery that maintains photoreceptors' organization and integrity, conferring them some protection against light induced damage. Our data point to a mechanism in which SFRP1 would favour the tight cellular adhesions underlying the OLM, which, in turn, is essential for photoreceptors' function and fitness.
In absence of Sfrp1, cones are the first photoreceptors to show abnormalities with misplaced cell bodies and occasional OS de-compaction detected since young adult stages and light-damage susceptibility. This is somewhat unusual because cones have been shown to be more resistant to this type of damage 57 . This unusual behaviour might reflect a particular dependence of cones on the interaction with the surround tissues 60 . Indeed, a number of genes responsible of cone dystrophies encodes proteins involved in cell adhesion 4,7,61 . Furthermore, overexpression studies in the embryonic chick retina indicate that Sfrp1 promotes cone generation 19 , although Sfrp1 −/− eyes show no developmental defects 18,30,31 . This may be because the related Sfrp2 and Sfrp5 compensate Sfrp1 function during development 12,18,32,62 . In the adult mouse retina, Sfrp2 and Sfrp5 mRNAs are absent or poorly expressed (Fig. S1), allowing to uncover a specific SFRP1 activity in the retina. We have observed that the initially subtle cone alterations worsen with time, and are followed by the appearance of rods' defects and visual function impoverishment. The emergence of these defects have been also observed in aged retinas 63,64 , in which, according to our data, SFRP1 content is progressively decreasing.
SFRP proteins have been initially discovered and described as apoptosis-related proteins 65 . We cannot fully discard that retinal SFRP1 may in part have such an activity. However, whereas SFRP2 and SFRP5 have anti-apoptotic functions 65,66 , SFRP1 over-expression increases cell sensitivity to pro-apoptotic stimuli 65 , making the defects observed in Sfrp1 −/− retina hardly interpretable through an apoptotic mechanism. Inhibition of Wnt signalling, a SFRP1 function observed in many different contexts 11 , is also unlikely in the adult retina, because photoreceptors are significantly less vulnerable to light-induced damage upon Wnt signalling activation 67 , which is the opposite of what we observe in Sfrp1 −/− mice. The Sfrp1 −/− retinal phenotype can instead be best explained considering the inhibitory function that SFRP1 exerts on the metalloprotease ADAM10 12,13 . In a speculative model (Fig. 8), we envisage that SFRP1, produced by photoreceptors and possibly Müller glial cells, down-regulates ADAM10 activity at the OLM, thereby maintaining the required tight adhesion between the end-feet of Müller cells and the IS of the photoreceptors. In its absence, the increased proteolysis of the OLM and (red arrows) and enlarged OS with evident breakage of the OLM (white arrows); (p) The graphs represent the thickness of the ONL in 12 months-old wt and Sfrp1 −/− retinas showing a significant reduction in the mutants more evident in central regions. Mann-Whitney U test; *p < 0.05, **p < 0.01. Abbreviations. bc: bipolar cell; hc: horizontal cell; r ribbon; sv: synaptic vesicles. www.nature.com/scientificreports www.nature.com/scientificreports/  . Sections were immunostained with antibodies against Rho (rods, red) or GNB3 (cones, green) and counterstained with Hoechst. (m,n) The graphs represent the ONL thickness (measures were taken in cryostat frontal sections of the eye at the level of the optic disk) in the different conditions. Kruskal Wallis with post hoc Dunn. *Indicates significant differences between Sfrp1 −/− untreated and light-damage mice. # Indicates significant difference between Sfrp1 −/− and wt animals exposed to light damage. No significant differences were found between wt untreated and light-damage mice. *p < 0.05, **p < 0.01. Scale bar in l applies to all panels. Abbreviations: ONL: outer nuclear layer; OPL: outer plexiform layer; OS outer segment. www.nature.com/scientificreports www.nature.com/scientificreports/ OS components loosen the OLM, progressively impairing photoreceptors' anchorage. The OLM confers mechanical strength to the retina and limits the diffusion of components of the photo-transduction cascade. Its loosening sets the conditions for loss of photoreceptors' fitness, making the cells more sensitive to light, and possibly, other types of stress (Fig. 8). This tentative model is supported by the appropriate localization of the required molecular machinery -ADAM10, SFRP1 and different cell adhesion molecules-within the retina 44,45,53,54 . Indeed, our data shows that both proteins localize in the OLM and regions of the INL compatible with Müller glial cell localization. We have previously shown that SFRP1 physically interacts with ADAM10 in both the retina and the telencephalon 12,13 , down-modulating the proteolysis of different ADAM10 substrates 12,13,22,25 . Consistent with this notion, here we have observed an increased shedding of N-cadherin and PCDH21 in the Sfrp1 −/− retinas. Increased proteolysis of these molecules seems to weaken cell-cell interactions 50,51 and thus may well explain the intermittent breakage of the OLM in the Sfrp1 −/− retinas, with the consequent disorganization of the ONL and rhodopsin distribution. The latter defect is considered both a consequence and a further cause of rod degenerative signs, including the decompaction of the OS disc membrane stacks 68 , as we observed in the Sfrp1 −/− mutants.
In further support for our findings, missense mutations in PCDH21 have been identified in patients affected by autosomal recessive cone-rod dystrophy 4,61 , suggesting that this protocadherin might be particularly relevant for cone survival. Indeed, its targeted disruption in mice causes OS fragmentation followed by a relatively mild decrease of the ERG response and photoreceptors' loss 52 , somewhat resembling the Sfrp1 −/− retinal phenotype. However, the precise relationship between PCDH21 shedding and photoreceptor degeneration might not be so straightforward. Analysis of PCDH21 shedding in mice that lack peripherin (rds −/− ), a structural OS protein, revealed higher levels of protein expression and a decreased proteolysis 52 and we show that in rd10 mice PCDH21 proteolysis was similar to wt. The reasons for these differences likely reflect the primary genetic causes of photoreceptors degeneration in rd10 and rds mice: PCDH21 and peripherin have structural functions whereas rd10 mice carry a mutation in a component of the rod phototransduction cascade 55 . Changes in cell adhesion might thus be a consequence rather a cause in rd10 mice.
Our study highlights a specific role of Sfrp1 in the adult retina. We have interpreted these findings in light of the idea that SFRP1 acts as an endogenous inhibitor of ADAM10 in other contexts 12,13,22,25 . Our findings however are in apparent contradiction with the observation that in the brain SFRP1 expression increases with aging 69 and is upregulated in individual affected by Alzheimer's disease 13 , whereas it is barely detectable in the cortex of young and healthy individuals. We believe that this apparent discrepancy is linked to differential distribution of the protein in  rods and rod bipolar cells, a,b) and photopic (cones and cone bipolar cells, c, d) ERG response from 4 months-old WT and Sfrp1 −/− animals to flash light stimuli of increasing brightness (cd/m 2 ). Kruskal Wallis with post hoc Dunn. *indicates significant difference between Sfrp1 −/− untreated and light-damage mice. #indicates significant difference between Sfrp1 −/− and wt animals exposed to light damage. + Indicates significant differences between Sfrp1 −/− and wt untreated mice. No significant differences were found between untreated and light-damage wt mice. *p < 0.05, **p < 0.01, ***p < 0.001. Abbreviations: ONL: outer nuclear layer; OPL: outer plexiform layer; OS outer segment. www.nature.com/scientificreports www.nature.com/scientificreports/ homeostatic conditions and the drastic alterations that occur during neurodegenerative events. In Alzheimer's disease, activated glial cells (particularly astrocytes) produce SFRP1, leading to increased levels of toxic products of APP processing 13 , which further sustain chronic inflammation 16 . In the adult retina, both neurons and likely Müller glial cells produce SFRP1, which is localized at the OLM with a putative homeostatic function over ADAM10. Its loss of function leads to subtle and slow alterations that are not associated with glial cell reactivity. Drastic photoreceptors' degeneration, leads instead to neuroinflammation [70][71][72] . We expect that in this condition Sfrp1 expression would be also up-regulated in reactive retinal glia, explaining the reported elevated expression and ectopic distribution of SFRP1 in RP 24,27 . We also expect that neuroinflammatory functions of SFRP1 16 would override its homeostatic role, although testing whether this is the case is not straightforward. SFRP1 expression is under strong epigenetic regulation 28 , which may further tune SFRP1 protective or offensive role in the retina.
In conclusion, we show that Sfrp1 contribute to adult retinal homeostasis maintaining the integrity of the OLM. The phenotype observed in Sfrp1 null mice suggest the possibility that the age-related decrease of retinal fitness and visual performance might involve a progressive down-regulation of Sfrp1 expression. Furthermore, differential SFRP1 expression among individuals may represent one of the many possible causes of the low genotype-phenotype correlation observed among individuals suffering of retinal dystrophies.
In situ hybridization (ISH) and immunohistochemistry (IH). Mice were anesthetized with sodium pentobarbital and perfused intracardially with 4% PFA in 0.1 M phosphate buffer (PB). Alternatively, animals were euthanized by CO 2 inhalation and eyes were enucleated and fixed by immersion in 4% PFA in PB for 2 h. The www.nature.com/scientificreports www.nature.com/scientificreports/ anterior segment of the eye was removed and the eyecups were cryoprotected by equilibration in 15-30% sucrose gradation in PB, embedded in 7.5% gelatine/15% sucrose solution, frozen and sectioned at 16 µm. Sections were processed for in situ hybridization (ISH) as described 22 using a DIG-labelled anti-sense riboprobe for Sfrp1, Sfrp2 and Adam10. When processed for immunohistochemistry (IH), the sections were incubated for 2 h at room temperature (RT) in a solution containing 1% BSA, 0.1% gelatine and 0.1% Triton X-100 in phosphate buffer saline (PBS). The sections were then incubated with primary antibodies diluted in blocking solution at 4 °C overnight. Primary antibodies were the following: rabbit anti-SFRP1(1:500), rabbit anti-GFP (Molecular Probes, 1:1000); rabbit anti-ADAM10 (1:500, Abcam); mouse anti-Rhodopsin K16-155C 74 (kindly provided by Dr. P. Hargrave, 1:100), rabbit anti-Guanine Nucleotide Binding protein (G protein Beta 3 subunit, GNB3; Abcam; 1:1000), rabbit anti-β-catenin (Abcam, 1:1000), mouse anti-N-cadherin (Invitrogen; 1:500), mouse anti-Glutamine Synthetase (GS; Millipore, 1:1000). Sections were washed and incubated with the appropriate Alexa-Fluor 488 or 594-conjugated secondary antibodies (Molecular Probes, 1:1000) at RT for 90 min. After washing, the tissue was counterstained with Hoechst (Sigma), mounted in Mowiol and analysed with light (Leica DM500) or confocal microscopy (Zeiss LSM 710). Confocal images were acquired with identical settings. Figures display representative images. Every staining was performed in a minimum of 3 control and 3 mutant mice, obtaining equivalent results. Retina thickness was measured using Hoechst stained cryostat sections at 400 to 1800 µm from the optic nerve head at intervals of 200 µm. Mean ± SEM thickness was calculated and plotted using Excel software. Mann-Whitney U test was applied for comparison between wt and mutant animals.

ELISA.
To quantify the amount of SFRP1 protein, retinas from WT mice were isolated and lysed in RIPA buffer. SFRP1 protein levels were determined using a specific capture ELISA as detailed in 13 . Transmission electron microscopy. Eyes were fixed by immersion in 2% glutaraldehyde and 2% PFA in 0.1 M cacodylate buffer. Samples were treated with 1% osmium tetroxide, dehydrated and embedded in Epon 812. Semi-thin sections (0.7 μm) were stained with toluidine blue and observed by light microscopy (Leica DM500), whereas ultrathin sections were analysed by electron microscopy (Zeiss EM900). Distributions of cone nuclei were measured in central retina semi-thin sections of 3 wt or mutant mice at (2 or 6 months old) and statistics was calculated and plotted using Excel software. Mann-Whitney U test was applied for comparison between wt and mutant animals. At least 3 wt or mutant mice were analysed using electron microscopy. Figures display representative images.
Western blot analysis. Isolated retinal tissue including the RPE was incubated in 2 mM CaCl 2 at room temperature for 30 min and then in buffer containing 150 mM NaCl, 50 mM Tris pH 7.5, 1 mM EDTA, 1% NP-40, complete protease inhibitor cocktail and PMSF (Roche), as previously described 52 . Total proteins (50 µg) were separated by SDS-PAGE and transferred onto PVDF membrane. Membranes were immersed in a solution of 5% non-fat milk and 0.1% Tween 20 in Tris Buffered Saline (TBS) for 2 h and then incubated with the appropriate primary antibodies diluted in blocking solution at 4 °C overnight. The primary antibodies used were: rabbit antiserum against the C-terminal domain of PCDH21 (a kind gift of Dr. A. Rattner; 1:10.000), mouse monoclonal antibody generated against the C-terminal domain of N-cadherin (Invitrogen; 1:500); anti-CRB1 (a kind gift of Dr. J. Wijnholds, 1:500), and mouse anti-α-tubulin (Sigma, 1:10.000), used as a loading control. The membranes were washed and incubated with horseradish peroxidase-conjugated rabbit or mouse antibodies followed by ECL Advanced Western Blotting Detection Kit (Amersham). The immune-reactive bands were quantified by densitometry and the amount of proteolyzed fragment was estimated as the ratio of C-terminal fragment (CTF)/full length (FL)/α-tubulin band values. Western blots were repeated at least 3 times obtaining similar results.
Light-induced retinal damage. The pupils from 4 and 9 months-old mice were dilated by topical administration of 1% tropicamide (Alcón cusí) 30 min before light exposure and by a second administration after the first 4 h of exposure. The light exposure device consisted of cages with reflective interiors, in which freely moving animals were exposed, at the same time of the day, to 15000 lux of cool white fluorescent light for 8 h. Control animals were sacrificed before light exposure, whereas experimental animals were sacrificed 7 days after completed light exposure.

Electroretinography (ERG).
Mice were dark-adapted overnight. All procedures were performed under dim red light following the ISCEV guidelines 75 . Animals were anesthetized by intra-peritoneal injection of ketamine (95 mg/Kg) and xylazine (5 mg/Kg). The body temperature was maintained with an electric heating pad. The pupils were dilated by topical administration of 1% tropicamide as above. A drop of 2% methylcellulose was then placed between the cornea and the silver electrode to maintain conductivity. VisioSystem Veterinary set up (SIEM Bio-Médicale) was used to deliver flash stimuli, as well as to amplify, filter and analyse photoreceptor responses. Rod responses (Scotopic ERG) were determined using stimulation intensities ranging from −2.8 to −0.2 log Cd s/ m 2 . Cone responses (Photopic ERG) were registered after 10 min adaptation to a 3 Cd/m 2 background light using stimulation intensities ranging from −0.35 to 0.35 log Cd s/m 2 and after 10 min adaptation to a 30.3 Cd/m 2 background light using a stimulation intensity of 0.35 log Cd s/m 2 . Rods may respond even in photopic conditions 76,77 . We thus tested cone response also by flicker ERG using different frequencies that allow discriminating cone from rod response. Indeed, rods only respond to low frequency stimuli (5-15 Hz) whereas cones can respond to both low and high frequencies   76,77 . Flicker ERG was performed with a stimulation intensity of 0.5 log Cd s/m 2 at frequencies of 10, 20 and 30 Hz, after 10 min adaptation to a 30.3 Cd/m 2 background light. ERGs were recorded in 4 months old mice, 5 wt and 6 mutant mice homozygotes for the protective Rpe65 Leu450Met gene variant, in 5 wt and 5 mutant mice heterozygotes for the protective Rpe65 Leu450Met gene variant and in 2 wt and 3 mutant mice without protective gene variant. ERGs were also recorded in 9 months old mice, 5 wt and 5 mutant