Deletion of the phosphatase INPP5E in the murine retina impairs photoreceptor axoneme formation and prevents disc morphogenesis

INPP5E, also known as pharbin, is a ubiquitously expressed phosphatidylinositol polyphosphate 5-phosphatase that is typically located in the primary cilia and modulates the phosphoinositide composition of membranes. Mutations to or loss of INPP5E is associated with ciliary dysfunction. INPP5E missense mutations of the phosphatase catalytic domain cause Joubert syndrome in humans—a syndromic ciliopathy affecting multiple tissues including the brain, liver, kidney, and retina. In contrast to other primary cilia, photoreceptor INPP5E is prominently expressed in the inner segment and connecting cilium and absent in the outer segment, which is a modified primary cilium dedicated to phototransduction. To investigate how loss of INPP5e causes retina degeneration, we generated mice with a retina-specific KO (Inpp5eF/F;Six3Cre, abbreviated as retInpp5e−/−). These mice exhibit a rapidly progressing rod–cone degeneration resembling Leber congenital amaurosis that is nearly completed by postnatal day 21 (P21) in the central retina. Mutant cone outer segments contain vesicles instead of discs as early as P8. Although P10 mutant outer segments contain structural and phototransduction proteins, axonemal structure and disc membranes fail to form. Connecting cilia of retInpp5e−/− rods display accumulation of intraflagellar transport particles A and B at their distal ends, suggesting disrupted intraflagellar transport. Although INPP5E ablation may not prevent delivery of outer segment–specific proteins by means of the photoreceptor secretory pathway, its absence prevents the assembly of axonemal and disc components. Herein, we suggest a model for INPP5E–Leber congenital amaurosis, proposing how deletion of INPP5E may interrupt axoneme extension and disc membrane elaboration.

INPP5E, also known as pharbin, is a ubiquitously expressed phosphatidylinositol polyphosphate 5-phosphatase that is typically located in the primary cilia and modulates the phosphoinositide composition of membranes. Mutations to or loss of INPP5E is associated with ciliary dysfunction. INPP5E missense mutations of the phosphatase catalytic domain cause Joubert syndrome in humans-a syndromic ciliopathy affecting multiple tissues including the brain, liver, kidney, and retina. In contrast to other primary cilia, photoreceptor INPP5E is prominently expressed in the inner segment and connecting cilium and absent in the outer segment, which is a modified primary cilium dedicated to phototransduction. To investigate how loss of INPP5e causes retina degeneration, we generated mice with a retina-specific KO (Inpp5e F/F ;Six3Cre, abbreviated as ret Inpp5e −/− ). These mice exhibit a rapidly progressing rodcone degeneration resembling Leber congenital amaurosis that is nearly completed by postnatal day 21 (P21) in the central retina. Mutant cone outer segments contain vesicles instead of discs as early as P8. Although P10 mutant outer segments contain structural and phototransduction proteins, axonemal structure and disc membranes fail to form. Connecting cilia of ret Inpp5e −/− rods display accumulation of intraflagellar transport particles A and B at their distal ends, suggesting disrupted intraflagellar transport. Although INPP5E ablation may not prevent delivery of outer segment-specific proteins by means of the photoreceptor secretory pathway, its absence prevents the assembly of axonemal and disc components. Herein, we suggest a model for INPP5E-Leber congenital amaurosis, proposing how deletion of INPP5E may interrupt axoneme extension and disc membrane elaboration.
In this study, we confirm that INPP5E localizes to the WT photoreceptor IS by multiple means. We also deleted INPP5E from the retina by mating Inpp5e F/F mice (12) with Six3Cre transgenic mice and observed that ret Inpp5e −/− rod outer segments (ROSs) initiate degeneration by postnatal day 10 (P10). While P10 connecting cilia (CC) of P10 mutant rods are nearly normal in length, shortened ROS are devoid of recognizable discs. Intraflagellar transport (IFT)-A and IFT-B particles accumulate in the proximal ret Inpp5e −/− OS, suggesting defective IFT, a bidirectional transport system powered by molecular motors (24). Thus, deletion of INPP5E primarily impairs axoneme extension and disc morphogenesis of both rods and cones. Inpp5e KO mice   INPP5E is a 72-kDa protein carrying a proline-rich domain  in the N-terminal region, a large phosphatase active site  encoded by exons 3 to 9, a coiled-coil domain, and a CAAX motif for C-terminal farnesylation (Fig. 1A). Mutations in human INPP5E associated with JBTS are located predominantly in the phosphatase domain. The mouse Inpp5e gene consists of 10 exons, producing two splice variants (Fig. 1B). The full-length variant is predicted to be farnesylated and membrane-anchored; the shorter variant lacking exon 10 and the CAAX box is predicted to be soluble.

Generation of retina-specific
To generate retina-specific KO ( ret Inpp5e −/− ) mice, Inpp5e F/ F mice were bred with Six3Cre mice (25). LoxP sites placed in introns 1 and 6 specify deletion of most of the phosphatase domain (Fig. 1C). LoxP, Six3Cre, and KO genotyping is shown (Fig. 1D). Immunoblotting with two independent antibodies against human INPP5E (Proteintech [PT]) and a mouse Nterminal fragment of INPP5E (custom-made at Covance, Inc. [CO]) confirmed the presence of two splice variants in WT retinal lysates and deletion of both isoforms in the conditional KO line (Fig. 1E).   (B) ERG traces at various intensities. C-H, scotopic and photopic ERG measurements as a function of light intensity conducted at P15 (C-E) and P21 (F-H). A-wave and b-wave amplitudes were plotted as a function of flash intensity. Note the significant photopic b-wave activity of the ret Inpp5e −/− retina at P15 (E) and minor haploinsufficiency at P21 (H). P15 control, n = 7 and KO, n = 6. P21 control, n = 3 and KO, n = 4. Data were analyzed via t test: two samples assuming equal variance. Data are presented as the mean ±SEM. All a-wave scotopic amplitudes are less than **p < 0.01, and all b-wave amplitudes are less than *p < 0.05. Mutant mice were compared with their respective WT and heterozygous littermates. P15 control, n = 7 and KO, n = 6. P21, WT n = 4, heterozygous littermates, n = 3, and KO n = 4. There is no statistical significance at P15 photonic b-wave amplitude at 1. 38

Electroretinography
We recorded whole-field scotopic (dark-adapted) and photopic (light-adapted) electroretinography (ERGs) at P15 and P21 at light intensities ranging from −1.63 to 2.38 log cd s/ m 2 (Fig. 2, A and B). Scotopic a-waves (Fig. 2, C and F) and scotopic b-waves (Fig. 2, D and G) of KO mice were barely recordable at P15 and P21. Scotopic a-waves of heterozygous littermates displayed normal a-wave amplitudes at P21 (Fig. 2F), indistinguishable from the WT, suggesting haplosufficiency. Surprisingly, the P15 ret Inpp5e −/− photopic ERG b-wave amplitudes were still recordable (Fig. 2E), presumably due in part to a slower degeneration of cones in the peripheral retina. At P21, ret Inpp5e −/− photopic ERG b-wave amplitudes were much reduced (Fig. 2H). The b-wave amplitudes of heterozygous littermates were somewhat reduced relative to the WT, but the reduction was statistically insignificant. ret Inpp5e −/− photoreceptors undergo rapid degeneration Plastic sections of ret Inpp5e −/− and littermate WT control retinas collected at P10, P15, and P21 (Fig. 3, A-F) confirmed rapid photoreceptor degeneration, revealed by a progressively shrinking outer nuclear layer (ONL). At P10, ret Inpp5e −/− retinas (Fig. 3D) showed only a slightly reduced ONL thickness compared with the WT (Fig. 3A), but the distance between the retinal pigmented epithelium and ONL was substantially reduced, suggesting attenuation of the IS and/or OS volume. At P15, nuclear tiers were reduced by 30% (Fig. 3E), and at P21, only one row of nuclei, presumably cones, was present ( Fig. 3F), consistent with rapid rd1-like degeneration. Evaluation of the ONL thickness across the entire retina (Fig. 3, G-I) showed 40% reduction in the central retina at P15 and nearly complete obliteration of the central ONL at P21. Persisting ONL thickness of the retinal periphery (Fig. 3I) is consistent with lower Cre expression in the peripheral retina under the Six3 promoter (25,26). The ret Inpp5e −/− inner retina, including the inner nuclear and ganglion cell layers, was practically unaffected as judged by near-normal morphology of P10, P15, and P21 cryosections (Fig. 3, D and E). To verify specificity of the human antibody in mouse, we generated an independent antibody (CO) against the N-terminal peptide (MPSKSASLRHTEAC) of mouse INPP5E. As its

Deletion of INPP5E in the retina
sequence is distinct from the human homolog, the antibody is predicted to be mouse-specific. This antibody showed strong immunoreactivity in the entire IS, along the outer limiting membrane, with traces in the ONL (Fig. 4E); immunoreactivity was erased when the antibody was saturated with the antigenic peptide (Fig. 4F). Anti-INPP5E immunoreactivity was still detectable in KO sections at P15 with the PT antibody ( Fig. 4B) and at P13 with the CO antibody (Fig. 4H), suggesting ineffective clearing of INPP5E or nonspecific binding. Differences in IS antigen localization using the PT antibody (fixed sections, Fig. 4A) and CO antibody (slightly fixed sections, Fig. 4E) are likely attributable to different fixation protocols (see Experimental procedures).
To further verify the IS localization of INPP5E, we performed the technique of combined serial tangential retina sectioning and immunoblotting (27). A Western blot of proteins from serial sections cut through a WT rat retina starting from the photoreceptor tips was probed with antibodies against rhodopsin (OS marker) and INPP5E (

Localization of INPP5E to the CC
Dissociated mouse rods (23) were probed with both PT and CO antibodies (Fig. 5, A-D). The OSs were identified by anti-CNGA1/3 (cyclic nucleotide-gated channel subunits) located in the OS cell membrane and the CC by anti-MAK (male germ cell-associated kinase) antibodies localizing to the CC (28). These results clearly showed the absence of INPP5E in the OS and its prominent presence in both the IS and CC. Another CC marker is centrin-2 (CETN2), a Ca 2+ -binding protein associated with the CC and basal body. To demonstrate localization of INPP5E at the CC in vivo, we perfused P10 Egfp-Cetn2 + ret Inpp5e +/− and ret Inpp5e −/− mice with 4% paraformaldehyde (PFA), postfixed overnight, generated 100-μm retina sections, performed antigen retrieval with 1% SDS, and analyzed the sections treated with anti-INPP5E antibody by confocal microscopy. The results show the presence of INPP5E specifically at the ret Inpp5e +/− CC (Fig. 5E) and the absence of INPP5E in ret Inpp5e −/− sections (Fig. 5F).

Absence of PDEδ does not affect INPP5E localization
The IS location of INPP5E was unchanged in cryosections obtained from Arl3 −/− mice (22), suggesting that it is not delivered in a complex with the prenyl-binding protein PDEδ. We verified that cellular distribution of farnesylated INPP5E is not dependent on PDEδ by analyzing its localization patterns in Pde6d −/− photoreceptors (Fig. 6). The OS of the PDE6D +/cones contain geranylgeranylated PDE6α' (cone PDE6 α-subunit) (Fig. 6A) and farnesylated GRK1 (G-protein receptor kinase) (Fig. 6B) whose transport to COS is completely impaired in the absence of PDEδ. (Fig. 6, D and E). In contrast, Rhodopsin and PDE6 in ret Inpp5e −/− retina In the WT and mutant retina at P6, the OS begin to expand and rhodopsin is present in nascent WT OS and smaller mutant OS (Fig. 7, A and B; P8 panels). At P10, mutant OSs are shorter than the corresponding WT OSs (Fig. 7, A and B; P10 panel) and rhodopsin begins to mislocalize and its CC (labeled with EGFP-CETN2) are located closer to the ONL, suggesting a shrinking, degenerating IS. At P12, the WT OSs are maturing, but the mutant OSs are stunted and rhodopsin accumulates throughout the ONL (Fig. 7, A and B, P12 panels). As judged by Western blotting (Fig. 7C, left panel), the level of rhodopsin in mutant rods is slightly reduced at P6 and about half at P10 and P12 (Fig. 7C, right panel).
Prenylated peripheral OS proteins (PDE6, GRK1) are thought to traffic the OS by diffusion, using PDEδ as a solubilization factor (29,30) and ARL3-GTP as a cargo dispensation factor (22,31,32). In WT and ret Inpp5e −/− at P6, trace amounts of PDE6 are present in the OS (Fig. 7, D and E). At P8 (just after onset of degeneration), the level of OS PDE6 in ret Inpp5e −/− rods is comparable with the WT rods, but it is significantly reduced at P10 (Fig. 7E), consistent with the reduction of their OS size. These results suggest that endoplasmic reticulum docking, for post-translational modifications of PDE6, is not affected by INPP5E ablation.

Mutant cones form spherical OSs
The COSs begin to form at P4 (33) and, in the WT retina, can be detected by S-opsin immunofluorescence at P6 (Fig. 8A). At P8 to P10, WT COSs extend (Fig. 8A) while KO COSs form spherical structures (Fig. 8B). At P12, WT COSs are nearly mature while KO COSs remain spherical and increase in size. Similar spherical structures were seen when retinal sections were probed with antibodies against ML-opsin, cone PDE6, GRK1, and GUCY2E (mouse retina guanylate cyclase or GC1) (Fig. 8, C-F). Electron microscopy revealed that the mutant COS does not form discs, is filled with vesicles of various sizes, and contains rudimentary fragments of axonemal extension. Yet, the CC appears to be normal in length (Fig. 8G). Mitochondrion (white arrow) at the distal IS confirms the cell is a cone.

Ultrastructure of the mutant ROS
We next explored the ultrastructure of WT and mutant rod photoreceptors using transmission electron microscopy. At P6, rods produce the primary cilium emanating from the basal body (Fig. 10A). The process of ciliogenesis is unaffected by the INPP5E KO (Fig. 10C). At P10, WT rods start forming the OS containing the regular disc structure (Fig. 10B). At this age, the CC structure of ret Inpp5e −/− rods remained normal and extended some axoneme fragments into the proximal OS, but a normal-size axoneme was not generated. Importantly, the membrane structure emanating from the CC was highly disorganized and failed to form an ordered stack of discs (Fig. 10D).
Immunolocalization conducted with retinal sections from WT and mutant mice (Fig. 11, A and B) demonstrated correct positions of CEP290 (centrosomal protein 290), RPGR (Retinitis Pigmentosa GTPase regulator), and SPATA7 (spermatogenesis-associated protein 7), suggesting that mutant CC is intact. Overall, the Inpp5e −/− CCs appears to be indistinguishable from WT CC.

Impairment of axoneme extension
IFT is a bidirectional protein transport pathway for proteins biosynthesized in the IS. IFT depends on molecular motors and IFT particles. We investigated the levels and positions of the IFT particles IFT88, IFT57, and IFT140 by immunostaining (Fig. 11C). IFT88 and IFT57 are subunits of the IFT-B particle, whereas IFT140 is a subunit of the IFT-A particle complex. Normally, IFT particles are strongly associated with  the basal body and proximal OS and are essential for supporting anterograde and retrograde transport of the ciliary material (35)(36)(37). In P12 WT photoreceptors, IFT88 and IFT57 were strongly stained next to the basal body and weakly to the proximal OS, whereas IFT140 was seen only at the basal body (Fig. 11C). In KO photoreceptors, IFT88, IFT57, and IFT140 massively accumulated at the CC/OS junction and in the proximal mutant OS (Fig. 11D, white arrows), suggesting impaired retrograde IFT.

Deletion of INPP5E in the retina
As Inpp5e −/− OS development is severely impaired, we next explored how far the axoneme could extend into these altered OS structures. In WT control at P10, the axoneme visualized with an anti-acetylated α-tubulin (Ac-Tub) antibody (red) was extended from the CC (green) for approximately 4 μm (Fig. 12,  A and C). In mutant rods, the axoneme was stunted and never exceeded 1 μm (Fig. 12, B and C), consistent with transmission electron microscopy results (Fig. 10D).

Photoreceptor cilia are distinct from the primary cilia
In this study, we show that INPP5E localizes to the photoreceptor IS and CC, but not to the OS (Figs. 4 and 5). This is in contrast to the primary cilia of hTert-retinal pigmented epithelium and IMCD3 cell where INPP5E populates the entire primary cilium (5,6,16). The major role of INPP5E in the primary cilia is to create a phosphoinositol-4-phosphate (PI4P)-rich ciliary membrane environment by catalyzing the conversion of PIP2 to PI4P, which is essential to regulate the ciliary localization of receptor proteins of the sonic hedgehog signaling pathway. Ciliary targeting of INPP5E in IMCD3 cells was shown to be PDEδ/ARL3 dependent (16). The absence of PDEδ or inhibiting PDEδ activity prevented INPP5E delivery to the cilium of mouse embryonic fibroblasts (17).
Using immunostaining with two different antibodies and serial tangential sectioning with immunoblotting (Fig. 4), we show that INPP5E is a predominantly IS protein. This is consistent with previous reports that INPP5E localizes to the Golgi apparatus in immortalized cell lines (2) and photoreceptors (22). Mouse Inpp5e expresses two splice variants, both presumably expressed in photoreceptors, with or without the C-terminal exon containing a CAAX box motif for isoprenylation. The long version of INPP5E is prenylated and membrane-associated, and the short version is predicted to be soluble and cytoplasmic. Both prenylated INPP5E and unprenylated INPP5E (C644A) were shown to localize to the Golgi in Tera-1 testis-derived cell lines, mediated by the N-terminal proline-rich domain (2). In mouse retina, prenylated INPP5E is present at lower levels in the ONL (Fig. 4, A, C, E and G), presumably docked to the endoplasmic reticulum for post-translational CAAX box processing (removal of -AAX from the CAAX box and carboxymethylation) (38). The bulk of INPP5E is localized throughout the entire IS. Importantly, INPP5E (most likely the prenylated form) also locates to the CC (Fig. 5, B, D and  E). Why and how INPP5E is excluded from the photoreceptor OS is unclear.
Inpp5e −/− OSs are stunted beginning at P8 and never reach the normal length. At P10, mutant photoreceptors start to degenerate, rhodopsin mislocalizes in the ONL, and the ONL thickness is reduced (Fig. 3). The mislocalization of rhodopsin at P10 and PDE6 at P12 and later (Fig. 7B) is likely a secondary effect of failed OS extension. Cone (Fig. 8G) and rod ret Inpp5e −/− photoreceptors (Figs. 10 and 11) form an apparently normal length CC but do not extend an axoneme into the OS and do not form discs (Figs. 8G, 11 and 12). Presence of the CC and absence of OS axoneme extensions and discs have been observed in a number of animal models lacking OS proteins, including rhodopsin KOs (39,40), rd1 mouse (PDE6b null allele) (41), and rds mouse (PRPH2 null allele) (42). Presence of CC and absence of OS were also observed in deletions of a number of ciliary proteins, including nephrocystin-5 (NPHP5 or IQCB1) (43), nephrocystin-1 (NPHP1) (44), AHI1 (jouberin) (45), lebercilin (LC5) (46), demonstrating that axoneme extension and disc morphogenesis is a complex process requiring a large number of components. Our analysis of Inpp5e −/− localization of several OS-resident proteins did not detect any trafficking abnormalities (Fig. 9), suggesting that INPP5E is dispensable for ciliary targeting of key OS proteins.

Phosphoinositides and tubby proteins
The biochemical function of INPP5E is to hydrolyze 5phosphates at the inositol ring of PIP2 phosphoinositides, generating PI4P (Fig. 13A). In photoreceptors, the concentration of phosphatidylinositol polyphosphates in RIS and ROS is very low at roughly 0.04 mol% (47). Recent measurements with fluorescent phosphoinositide sensors revealed that the majority of PI4P and PIP2 in rods is present in the IS and cytoplasm surrounding the nuclei (48); the OS also contains smaller amount of PI4P detectable by MS but no detectable PIP2. The latter is consistent with a direct measurement showing the presence of trace amount of PI4P but not PIP2 in the isolated bovine OS (49). Our results, specifically the localization of INPP5E at the CC (Fig. 4J) and misplacement of PIP2-interacing TULP1/3 to the mutant OS (Fig. 9, I and J), suggest an involvement of phosphoinositides in axoneme elongation and disc morphogenesis.
Tubby family proteins (Tubby, Tulp1-4) are known to bind PIP2 through their C-terminal "tubby domain" (50). TULP1 was shown to be involved in transport of rhodopsin and is thought to interact with F-actin in photoreceptors (51,52). TULP3 is required for primary ciliary entry of ARL13b and INPP5E (20) and for ciliary entry of certain GPCRs by bridging the IFT-A complex with PIP2 (53,54). The increase of TULP3 may explain the accumulation of IFT proteins at the OS base (Fig. 11D) as TULP3 also binds to IFT-A (53,54). Accumulation of PIP2, TULP3, and IFT particles was also observed in the Inpp5e −/− primary cilia of neural stem cell cultures with ciliary defects (5,6).

Model of INPP5E-Leber congenital amaurosis
We propose a model in which INPP5E is located at the inner leaflet of the CC membrane (Fig. 13B) where it controls the composition and levels of PI4P and PIP2 phosphoinositides, as has been observed in primary cilia (5,6). In WT CC and proximal OS, the concentration of PIP2 is very low due to the activity of INPPP5E. Deletion of

Deletion of INPP5E in the retina
INPP5E is expected to increase the concentrations of PIP2 produced by a PI4P-5K kinase. PIP2 infiltrates the CC and accumulates in the proximal OS. Excess PIP2 attracts TULP3 and TULP1 and causes accumulation of retrograde IFT particles at the OS base, thereby causing an IFT impairment. Because anterograde IFT, mediated by kinesin-2 (36), and retrograde IFT, mediated by dynein-2 (55,56), are essential for OS formation and extension, KO OSs fail to extend their axonemes and form discs. Open questions are whether axoneme extension is necessary for disc formation and whether Tubby protein translocation to the KO OS directly impairs IFT. Future research will explore these questions and determine the levels of PIP2 at the CC and the OS base. It will be essential to further characterize changes in protein composition at this location and, ultimately, determine the exact mechanism preventing disc morphogenesis and axoneme extension.

Animals
Procedures were approved by the University of Utah Institutional Animal Care and Use Committee and were conducted in compliance with the National Institutes of Health Guide for Care and Use of Laboratory Animals. Floxed Inpp5e mice (Inpp5e F/F ) were provided mice (Dr Chris Mitchell, Monash University, Australia) and maintained in a 12:12 h dark-light cycle. A transgenic mouse expressing EGFP-CETN2 fusion protein (JAX stock # 008234) was used to identify centrioles/ transition zones (57). Three-month-old Long-Evans rats used for tangential retinal sectioning were obtained from Charles River.

Immunoblotting
For Western blotting, two retinas from one mouse were homogenized in 100 μl of 50-mM Tris HCl (pH 8), 100-mM NaCl, 10-mM EDTA, and 0.2% Triton X-100, and 2 μl of 100-mM PMSF and 2-μl protease inhibitor. Samples were then sonicated for 2 x 20 pulses at an intensity of 30% and spun at 15,000 rpm for 20 min. The protein concentration was determined by the Bradford assay. Proteins of retina lysates were separated by 10% SDS-PAGE and transferred to a nitrocellulose membrane (59). The resulting membrane was sequentially subjected to blocking for 1 h, primary antibody incubation overnight at 4 C, and secondary antibody incubation for 1 h. Primary antibodies were diluted 1:500 for anti-INPP5E (PT) and 1:5000 for anti-Rho (4D2). Secondary antibodies were iR680 goat-anti-mouse (1:5000) and iR800 goat-anti-rabbit (1:3000) (Odyssey); images were acquired using an Odyssey scanner. The intensities of protein bands were measured using ImageJ and normalized to control bands.

Confocal immunohistochemistry
Animals were dark-adapted overnight and sacrificed in ambient light. For conventional fixation (60), P6-P21 eyes from control and mutant mice were immersion-fixed using freshly prepared 4% PFA in 0.1 M phosphate buffer, pH 7.4, for 2 h on ice. Eyecups were then moved to 15% sucrose in phosphate buffer for 1 h and then to 30% sucrose overnight for cryoprotection.
For the low fixation protocol (CO INPP5E antibody and MAK CC marker do not work with standard fixation methods), mouse eyes were rinsed in 1x PBS followed by embedding in optical cutting temperature (OCT). Blocks were immediately frozen at −80 C and sectioned at 14 um. Slides were removed from freezer and warmed no longer than 5 min. 1% PFA (made by diluting 4% PFA in PBS) was applied to the slide for 2 min. Slides were washed for 5 min in 1x PBS. Sections were blocked using 10% normal goat serum in 0.1 M phosphate buffer-0.1% Triton X-100 for 1 h and incubated with the primary antibody overnight in a rotating humidified chamber at 4 C.
To show INPP5E localization at the CC in vivo, deeply anesthetized P10 mice were transcardially perfused with 7-ml 4% PFA in 0.1 M PBS, pH 7.4, at room temperature (RT) and enucleated eyes were postfixed overnight at 4 o C. Samples were cryoprotected in 30% sucrose, embedded in optical cutting temperature, and cut into 100-μm sections. Free-floating

Deletion of INPP5E in the retina
Retinal serial sectioning with Western blotting Experiments were performed as described (61,62). Rats were sacrificed, and eyes were enucleated and dissected in icecold Ringer's solution. A retina punch (3 mm diameter) was cut from the eyecup with a surgical trephine positioned adjacent to the optic disc, transferred onto PVDF membrane with the photoreceptor layer facing up, flat-mounted between two glass slides separated by plastic spacers (ca. 240 μm), and frozen on dry ice. The retina surface was aligned with the cutting plane of a cryostat, and uneven edges were trimmed away. Progressive 5-μm tangential sections were collected, subjected to SDS-PAGE, and probed with antibodies against INPP5E (PT) and rhodopsin (4D2).

Measurement of the ONL thickness
Mouse eyecups, with the anterior segments and lens removed, were fixed overnight in 4% PFA in PBS overnight. Eyes from both control and mutant mice were then immersed in 15% sucrose in the phosphate buffer for 1 h and then to 30% sucrose overnight for cryoprotection. Twelve-micron transverse sections were stained with 4 0 ,6-diamidino-2-phenylindole, and the thicknesses of the ONL layers were measured along the retinal vertical meridian at approximately 500 μm apart on each side of the optic nerve head.
After dark adaptation overnight, mice were anesthetized by intraperitoneal injection of ketamine and xylazine (0.1 mg and 0.01 mg/g body weight, respectively), and pupils were dilated with 1% tropicamide. Full-field scotopic retinal electrical responses were recorded in the dark upon a series of white-light stimuli shown below at increasing intensities. Full-field photopic retinal electrical responses were subsequently recorded after light adaptation for 5 min at 33.25 cd s/m 2 . The a-wave amplitude of scotopic ERGs was measured from the baseline to the peak of the cornea-negative wave, and the b-wave amplitude of scotopic and photopic ERGs was measured from the peak of the cornea-negative wave to the peak of the major cornea-positive wave.

Transmission electron microscopy
Isolated mouse eyecups at ages of P8 and P10 were fixed by immersion in fixative (1.5% glutaraldehyde-1% PFA in 0.1 M cacodylate buffer, pH 7.4) at 4 C overnight (32, 46). The eyecups were postfixed with 1% osmium tetroxide in 0.1 M cacodylate for 1 h, buffer-rinsed, stained en bloc with uranyl acetate, and subsequently dehydrated in an ascending series of methanol solutions. Eyecups were embedded in Epon resin (Ted Pella, Inc) for sectioning. 1 μm plastic sections were cut to face and orient photoreceptors near the optic nerve. Retina ultrathin (60 nm) sections were cut and transferred onto slot grids with carbon-coated Formvar film (Electron Microscopy Sciences) and poststained with uranyl acetate followed by lead citrate. Transmission electron microscopy was performed at 75 kV using a JEOL electron microscope.

Statistics
SigmaPlot12 was used for statistical analysis using student t test and the level of statistical significance was set p = 0.05.

Data availability
All data contained in the manuscript are located within the article.