Deletion of both centrin 2 (CETN2) and CETN3 destabilizes the distal connecting cilium of mouse photoreceptors

Centrins (CETN1–4) are ubiquitous and conserved EF-hand–family Ca2+-binding proteins associated with the centrosome, basal body, and transition zone. Deletion of CETN1 or CETN2 in mice causes male infertility or dysosmia, respectively, without affecting photoreceptor function. However, it remains unclear to what extent centrins are redundant with each other in photoreceptors. Here, to explore centrin redundancy, we generated Cetn3GT/GT single-knockout and Cetn2−/−;Cetn3GT/GT double-knockout mice. Whereas the Cetn3 deletion alone did not affect photoreceptor function, simultaneous ablation of Cetn2 and Cetn3 resulted in attenuated scotopic and photopic electroretinography (ERG) responses in mice at 3 months of age, with nearly complete retina degeneration at 1 year. Removal of CETN2 and CETN3 activity from the lumen of the connecting cilium (CC) destabilized the photoreceptor axoneme and reduced the CC length as early as postnatal day 22 (P22). In Cetn2−/−;Cetn3GT/GT double-knockout mice, spermatogenesis-associated 7 (SPATA7), a key organizer of the photoreceptor-specific distal CC, was depleted gradually, and CETN1 was condensed to the mid-segment of the CC. Ultrastructural analysis revealed that in this double knockout, the axoneme of the CC expanded radially at the distal end, with vertically misaligned outer segment discs and membrane whorls. These observations suggest that CETN2 and CETN3 cooperate in stabilizing the CC/axoneme structure.

Although it is well-established that centrins in lower eukaryotes are required for centriole replication and positioning (6)(7)(8), the requirement of centrins during vertebrate centriole duplication is controversial (9 -14). A recent study found that CETN2 positively regulates primary ciliogenesis by removing the centriole cap protein CP110 in vitro (15). Morpholino-based depletion of Cetn2 in zebrafish embryos also leads to cilia loss in multiple tissues (16).
There are four centrin genes (CETN1-4) in mammalian genomes that encode the Vfl2 (Chlamydomonas centrin)-like CETN1, -2, and -4 and the budding yeast CDC31-related CETN3 (17,18). CETN1 is highly expressed in sperm and ciliated cells (19). CETN2 and CETN3 are expressed ubiquitously in all somatic cells (4,20), and CETN4 is expressed in ciliated tissues (21). Knockout of Cetn1 in mouse causes male infertility because of centriole rearrangement deficiency at a late stage of spermiogenesis (22). Knockout of Cetn2 in mouse also leads to dysosmia and hydrocephalus as a result of impaired olfactory ciliary trafficking of adenylate cyclase III (ACIII) and cyclic nucleotide-gated channel ␣2 (CNGA2) and disrupted planar polarity of ependymal cilia, respectively (23). In germline knockouts of CETN1 and CETN2 and in CETN1/CETN2 double knockouts, rod and cone photoreceptors develop normally and exhibit normal function (23). A possible explanation for this is the functional centrin redundancy. CETN1-3 localize to the lumen of the axoneme of the photoreceptor CC (24,25), which connects the inner segment (IS) to the outer segment (OS). Each connecting cilium is an elongated TZ of ϳ1.1 m length (compared with 0.2-0.5 m in primary cilia) consisting of an array of nine microtubule doublets emanating from the basal body. It was recently found that the CC can be divided into two domains, a proximal CC corresponding to the TZ of primary cilia and a distal photoreceptor-specific CC that supports the very large photosensitive OS (26).
Our results show that removal of both CETN2 and CETN3 from the CC lumen leads to a widening of the distal CC microtubule doublet array similar to observations in a Spata7 knockout (26). CETN1 accumulates at the mutant CC center, presumably stabilizing the proximal CC. Weakening of the distal CC axoneme leads to misalignment of discs, formation of membrane whorls, OS malformation, and progressive photoreceptors degeneration.

Generation of Cetn3 KO mouse
The mouse Cetn3 gene consists of 5 exons. We generated gene-trapped Cetn3 mice (Cetn3 GT/GT ) in which FRT sites flanked the gene trap cassette inserted into intron 2 and LoxP sites flanked exon 3 (Fig. 1A). The presence of the gene trap truncates CETN3 after amino acid 52, producing a nonfunctional N-terminal fragment containing the first EF-hand motif (EF1). Correct gene targeting was confirmed by PCR amplification of 5Ј and 3Ј recombination arms, in addition to FRT and LoxP sites and DNA sequencing (Fig. 1, B and C). Cetn3 GT/GT mice (Fig. 1D) were crossed with Flp mice to remove the GT cassette, generating floxed mice (Cetn3 f/f ), which were subse- . The En2SA-IRES-LacZ-pGK-Neo GT cassette, which is flanked by two FRT sites (green triangles), is inserted into intron 2 by homologous recombination. Exon 3 is flanked by two LoxP sites (red triangles). B and C, PCR genotyping of 3Ј and 5Ј recombination arms (B) and 1st FRT, 2nd LoxP, and 3rd LoxP sites in two ES cell clones, B09 and E12 (C). D and E, PCR genotyping result of one litter each of Cetn3 GT/ϩ ϫ Cetn3 GT/ϩ (D) and Cetn3 ϩ/Ϫ ϫ Cetn3 ϩ/Ϫ (E) pups. Primer sequences are listed in Table 2. F and G, immunohistochemistry of WT (F) and Cetn3 GT/GT (G) retina cryosections incubated with anti-CETN3 antibody. Signal attributable to CETN3 is present in CC and basal bodies of WT photoreceptors (left) but absent from Cetn3 GT/GT photoreceptors (right). Scale bar: 5 m.

Centrins and retina degeneration
quently mated with a CMV-Cre line to delete exon 3 (Fig. 1E), thus truncating CETN3 in exon 4 in a different frame. Both the Cetn3 GT/GT and Cetn3 Ϫ/Ϫ mice were born in a Mendelian ratio (29 knockouts of 116 total pups (Table 1)), were viable and fertile and did not reveal syndromic ciliopathy. As reported (24), CETN3 protein localized to the wildtype (WT) photoreceptor CC and both centrioles (Fig. 1F). In Cetn3 GT/GT (Fig. 1G) and Cetn3 Ϫ/Ϫ (not shown) photoreceptors, CETN3 was undetectable, thereby confirming the null allele in each line.

Cetn2 ؊/؊ ;Cetn3 GT/GT mice exhibit progressive retina degeneration
Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT pups survived to adulthood and showed syndromic ciliopathy, as observed in Cetn2 Ϫ/Ϫ mice, including dysosmia and hydrocephalus (not shown). To identify a retina phenotype, we examined Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT retina morphology at 1, 3, and 13 months with anti-GC1 mAb (an OS marker) and DAPI (4Ј, 6-diamidino-2-phenylindole) as a nuclear marker (Fig. 3). We found that at 1 month, the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT ONL thickness and OS length were comparable with that of WT controls in dorsal and ventral retina (Fig. 3A, first and second column). By 3 months of age, the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT ONL thickness and OS length were greatly reduced, and the phenotype was more severe in the dorsal retina than in the ventral retina (Fig. 3A, third column). By 13 months, only one layer of ONL nuclei remained at the dorsal retina with residual OS/IS, as compared with 4 -5 layers of nuclei at the ventral ONL. The ventral retina was much more stable, displaying OS of half-normal length (Fig. 3A, fourth column). We found a consistent reduction of ONL thickness and OS lengths across 3 month-old Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT retinas (Fig. 3, B and C). The Cetn3 GT/GT central retina displayed about 10 rows of nuclei (45-50-m thickness) at either the dorsal or ventral areas, whereas the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT retina had 6 -7 rows of nuclei (30 -35 m) at the ventral retina and 4 -5 rows of nuclei (20 -25 m) at the dorsal retina (Fig. 3B). Cetn3 GT/GT central retinas had an average OS length of 15-25 m (both dorsal and ventral), whereas the OS length of the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT ventral retina was reduced to 10 -15 m and the dorsal peripheral retina to 5-10 m (Fig. 3C). At 1 month, the ONL thickness and OS length of Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT was comparable with the Cetn3 GT/GT control at both the dorsal and ventral retina (Fig. 3, D and E). These results show that lack of CETN2 and CETN3 caused a slow photoreceptor degeneration beginning after ϳ1 month of age and nearing completion in the dorsal retina at 1 year.

Centrins and retina degeneration OS protein trafficking proceeds in the absence of CETN2 and CETN3
Immunolabeling with antibodies directed against phototransduction proteins showed normal localizations of rod or cone OS proteins (rhodopsin, GC1, PDE6, CNGA1/A3, and Sand M-opsins) in 3-month-old Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT retinas despite ongoing photoreceptor degeneration (Fig. 5, A and B). A notable observation was the enlargement of cone OS in Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT mice (arrows in Fig. 5B); WT cone outer segments labeled with anti-cone arrestin taper distally (Fig. 5C), whereas a fraction of the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT cone OS was swollen to triple diameter (Fig. 5D). The swelling (up to 3 m in diameter) was even more obvious when sections were immunolabeled with anti-M-opsin (Fig. 5E).

CETN1 accumulation in the Cetn2 ؊/؊ ;Cetn3 GT/GT central CC
Double immunolabeling with anti-CEP290 and anti-CETN1 antibodies showed a nearly full overlap of CEP290 and CETN1 along the entire P25 WT CC (98%, Fig. 6, H-J and T). In P25 Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT photoreceptors, CEP290 labels the fulllength CC, but CETN1 appears to accumulate at the CC center with variable depletion occurring at the proximal and distal

Centrins and retina degeneration
both CETN2 and CETN3, CETN1 is gradually depleted from the CC distal and proximal ends and accumulates in the center, concurrent with a slight shortening of the Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT CC.

Misaligned Cetn2 ؊/؊ ;Cetn3 GT/GT OS discs and dilated CC/axonemes
Ultrastructure of 2.5-month-old Cetn3 GT/GT photoreceptors revealed normal, densely packed, vertically oriented rods with horizontally stacked OS discs (Fig. 8A). By contrast, the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT littermate OS structure was highly disorganized, as judged by overgrown and longitudinally aligned discs, membrane whorls, and expanded OS diameters (Fig. 8,  B-D), resembling the phenotype seen in Rp1 Ϫ/Ϫ or Rp1 knockin mouse retinas (28,29). There were also CC/axoneme structural abnormalities in the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT photore- . Ac-tubulin colocalize with CEP290 at CC (arrowheads) but Ac-tubulin signal also extends into and peaks at the proximal axoneme (arrows) in P25 and 3-month-old WT photoreceptors. In P25 Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT , the pattern is similar. By 3 months, the Ac-tubulin signal at the CC is reduced (arrowheads in J-L), whereas the signal at the proximal axoneme is increased (arrows in J-L). Scale bar: 2 m. M and N, representative Ac-tubulin (red) and CEP290 (green) signal profile along the CC-axoneme axis, starting from the CC proximal end of the WT (top panels) and Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT (bottom panels) at P25 (M) and 3m (N). Note the peak of the Ac-tubulin signal at 1.2-1.4 m with the sharp drop of CEP290 signal. O, quantification of Ac-tubulin signal ratio at CC versus proximal axoneme. The ratio was calculated by comparing the signal at the midpoint of the CC versus the signal peak of the OS axoneme base (measuring positions are marked by arrowheads and arrows). No difference was detected for the P25 samples, but at 3m the ratio is significantly lower in Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT than in WT. Shown are mean Ϯ S.D.; n ϭ 35-59 for each group, one-way ANOVA; ***, p Ͻ 0.001.

Centrins and retina degeneration
ceptors. Control Cetn3 GT/GT CC and axonemes had a normal diameter of ϳ250 nm (Fig. 8E). However, ϳ20 -30% of the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT photoreceptors (depending on the mouse), of Ͼ60 CC that were examined from three samples, showed variable dilation at the distal CC (Fig. 8, E-H, yellow double arrows) and OS proximal axoneme (Fig. 8I, red arrowheads) not seen in WT controls. This phenotype ranged from a minor, slight expansion (Fig. 8G) to severe, extreme dilation with invasion of vertically aligned disc membranes (Fig. 8H) and loss of microtubule doublet integrity (not shown). Consistent with CEP290 immunolabeling (Fig. 6, N-S and W), the CC average length (Fig. 8, E and F, red double arrows) was slightly but significantly reduced in 2.5-month-old Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT photoreceptors (0.87 Ϯ 0.14 m) compared with Cetn3 GT/GT controls (1.01 Ϯ 0.15 m) (Fig. 8J). These observations suggest that CETN2 and CETN3 are required for CC length and structural maintenance, which direct photoreceptor OS disc assembly.

Centrins and retina degeneration
Cetn3 GT/GT retinas (Fig. 11A). Although altered ciliary tubulin acetylation occurs in Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT , we did not observe any significant Ac-␣-tubulin level change (Fig. 11A). Distributions of peripherin 2 to the OS (Fig. 11, B and C) and prominin 1 to the OS base (Fig. 11, D and E) were preserved in Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT as in the Cetn3 GT/GT control retina. In addition, the intraflagellar transport protein IFT88 correctly localized to the basal body and OS axoneme base as two pools (basal body and proximal OS) in both Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT and control photoreceptors (Fig. 11, F and G) and as reported (33).

Discussion
We generated Cetn3 GT/GT and Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT mouse lines to investigate the function of CETN2 and CETN3 in pho-toreceptors. We had observed previously the absence of a retinal phenotype in Cetn1 Ϫ/Ϫ and Cetn2 Ϫ/Ϫ mice (22,23), suggesting centrin redundancy. Here, we found that 1-year-old Cetn3 GT/GT single knockouts exhibited completely normal retina morphology and function. However, Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT mice revealed progressive retina degeneration starting at 1 month of age, which was nearly complete in the dorsal retina 1 year later. Our main results were as follows: (i) Cetn2 ϩ/Ϫ ; Cetn3 GT/GT mice display reduced scotopic a-and photopic b-waves, signaling haploinsufficiency (one allele of Cetn2 is insufficient to establish the WT phenotype); (ii) Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT mice are born in a non-Mendelian ratio, suggesting loss of embryos during embryonic development; (iii) Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT photoreceptor CC are shortened and dilated dis-

Centrins and retina degeneration
tally; and (iv) SPATA7 depletes and CETN1 concentrates in the mid-segment of the mutant CC (Fig. 12). The results demonstrate that CETN2 and CETN3 together stabilize the photoreceptor distal CC and proximal OS axoneme by interacting with the protein complex of which SPATA7 is a member.
Unlike their orthologs in lower eukaryotes (34 -36), the requirement for vertebrate centrins in centriole duplication (9,(11)(12)(13)(14) or ciliogenesis (15,23) is still controversial with inconsistent results among different studies. Although we have not directly assessed centriole duplication, the lack of detectable phenotype in Cetn3 GT/GT mice indicates that CETN3 is dispensable for centriole duplication in mouse, in contrast to a previous study with Xenopus embryos in which ectopic recombinant human CETN3 protein inhibits centriole duplication and blastomere cleavage (14). As Cetn2 Ϫ/Ϫ pups are born slightly below a Mendelian ratio (Table 1), CETN2 is probably required for centriole duplication in a small percentage of mouse embryos (23). However, the severely reduced Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT pup ratio (29.2% versus theoretical 50% (Table 1)) indicates that normal mouse embryo development requires both CETN2 and CETN3, implying that CETN2 and CETN3 are both required for centriole duplication and mitosis during embryonic development. Why some embryos can escape CETN2/CETN3 deficiency but others cannot and why only selected tissues show ciliopathy in Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT mice, despite the ubiquitous expression of CETN2 and CETN3, is unknown. Incom-

Centrins and retina degeneration
plete penetrance is often observed in ciliopathy, caused by basal body gene mutations (37).
Given centrin localization in the basal body and connecting cilium lumen, photoreceptors are an excellent system to use for studying the roles of CETN2 and CETN3 in postmitotic neurons. Our results show that CETN2 and CETN3 are together required for photoreceptor survival after maturation. In the absence of CETN2 and CETN3, CC and axonemes display molecular and structural changes, including reduction of the RP1 decoration of proximal OS axonemes (Fig. 6), CETN1 accumulating at the center of the CC but depleted from the distal and proximal ends (Fig. 6), microtubule hypoacetylation at the CC but hyperacetylation at the proximal OS axoneme (Fig. 7), SPATA7 depletion from the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT CC (Fig. 9), CC shortening (Figs. 6 and 8), and distal CC and proximal OS axoneme expansion (Figs. 8 and 12). RP1 is a photoreceptor-specific axoneme-and microtubule-binding protein. RP1 knockout, or knockin of a mutant form, causes progressive retina degeneration and disc morphogenesis defects (28,29). Notably, Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT mice showed OS malformation and disorganization very similar to that seen in RP1 mutants, i.e. vertically misaligned discs, membrane whorls, and disc expansion (Fig. 8), which could occur in RP1 mutants as early as P7 when rod OS discs first assemble (28,29). Apart from OS disc orientation, RP1 is also important in controlling the length and stability of the photoreceptor axoneme (27). Despite these similarities, noticeable phenotypic differences do exist in Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT mice relative to RP1 mutants. First, the photoreceptor degeneration is slower in the Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT retina. RP1 retina degeneration is nearly complete by 10 months (28,29), but in 13-month-old Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT mice, the ventral retina still has 4 -5 rows of ONL nuclei (Fig. 3A). Second, faster degeneration of the dorsal versus the ventral retina is unique to Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT mice (Fig.  3A). Third, dilation of the CC and axoneme (Fig. 8) and extreme swelling of the Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT cone OS (Fig. 5, C-E) have not been reported for RP1 mutants.
Ciliary tubulin hypoacetylation, proximal OS axoneme tubulin hyperacetylation, and CETN1 depletion from the CC distal ends in Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT photoreceptors are intriguing observations. Tubulin acetylation is a hallmark of long-lived, stable microtubules. Mice lacking ␣TAT1 (␣-tubulin acetyltransferase 1), the enzyme predominantly responsible for ␣-tubulin acetylation in vivo, are viable and display no overt phenotypes except for the deformation of the dentate gyrus (38). Whether a TAT1-deficient mouse has a retina phenotype has not been reported. Recently, it was found that intraluminal tubulin acetylation protects microtubules from mechanical breakage by weakening the lateral interaction of protofilaments and enhancing microtubule flexibility (39,40). As microtubule bending occurs in response to the mechanical force generated by microtubule motor movement (41) and actomyosin contractility (42,43), we predict that a consequence of CC hypoacetylation would be reduced resistance to microtubule bending and increased probability of breakage at CC. This notion is supported by the observation of the expansion/break of the distal CC and proximal axoneme ( Fig. 8 and not shown). How centrins regulate tubulin acetylation is largely unknown. Centrins are localized within the CC lumen along the microtubule surface (24,25), but ␣-tubulin acetylation and deacetylation occur on Lys-40, located at the microtubule intraluminal side. Thus, centrins may indirectly regulate Lys-40 acetylation by ␣TAT1 (44) or deacetylation by HDAC6 (histone deacetylase 6) (45), and one possibility is that centrins regulate the entry of ␣TAT1 or HDAC6 into the microtubule lumen.
The photoreceptor CC has long been considered analogous to the TZ of the prototypic primary cilium, but a recent study shows that the photoreceptor distal CC (DCC) is uniquely maintained by a retina-specific ciliopathy protein, SPATA7 (26). Common TZ proteins (such as NPHP1, NPHP4, NPHP6, AHI1, RPGR, and RPGRIP) are lost specifically in Spata7 Ϫ/Ϫ DCC but not the proximal Spata7 Ϫ/Ϫ CC of mouse photoreceptors, which collectively cause microtubule destabilization and axoneme radial expansion (26). In the Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT mouse, SPATA7 is gradually depleted from the CC (Fig. 9), which provides a mechanistic explanation for the DCC/axoneme dilation phenotype. Interestingly, although SPATA7 is nearly completely depleted from the 3m Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT photoreceptor CC, RPGR, RPGRIP (Fig. 10), and NPHP6 proteins (Figs. 6 and 7) are positioned normally (RPGR and NPHP6) or affected minimally (RPGRIP). These observations suggest that loss of transition zone proteins, including RPGR and RPGRIP, may not fully account for the DCC microtubule phenotype seen in Spata7 Ϫ/Ϫ mouse. Alternatively, as a microtubule-associated protein (46), SPATA7 may stabilize microtubules directly. Notably, CETN2 specifically is lost from the DCC of Spata7 Ϫ/Ϫ photoreceptors (26). DCC dilation is also observed in the Fam161a knockout mouse, a model of human RP28 characterized by CETN3 exclusion from the CC (47). We observed that CETN1 is depleted from the DCC and, to a lesser extent, from the proximal CC in Cetn2 Ϫ/Ϫ ; Figure 12. Model of CETN2 and CETN3 function in mouse photoreceptors. In WT photoreceptors, CETN1-3 are localized to the CC and the centriole lumen formed by the inner microtubule wall. SPATA7 is located outside along the outer CC microtubule wall. The axoneme (diameter of ϳ250 nm) extends into the OS where discs are formed. In Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT photoreceptors, CETN1 accumulates at the CC center and SPATA7 is gradually depleted from the CC outside wall. As a consequence, distal CC and basal OS axonemes are destabilized, the microtubule structure disintegrates, and OS discs misalign; BB, basal body; DC, daughter centriole.

Centrins and retina degeneration
Cetn3 GT/GT photoreceptors (Fig. 6). Centrins are not only perfect molecular markers for CC but also actively participate in maintaining connecting cilium and proximal axoneme structural stability. How CETN2 and CETN3 interact specifically and regulate SPATA7 localization warrants further investigation.
In summary, CETN2 and CETN3 have redundant functions in vivo. Our Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT mouse is a valuable model for studying centrin function in centriole duplication and ciliogenesis. CETN2 and CETN3 are both required for photoreceptor survival by regulating SPATA7 CC localization, DCC/axoneme microtubule stability, and OS disc orientation (see model in Fig. 12).

Animals
Mouse procedures were approved by the University of Utah Institutional Animal Care and Use Committee (IACUC) and were conducted in compliance with the NIH Guide for Care and Use of Laboratory Animals.

Generation of Cetn3 GT/GT and Cetn3 ؊/؊ mouse
Cetn3 embryonic stem (ES) cells (clones EPD0630_2_E12 and_ EPD0630_2_B09) on a C57BL/6N genetic background, JM8A3.N1 subline) and containing a gene trap (GT) in intron 2, were acquired from EUCOMM (Helmholtz Zentrum, Munich, Germany). ES cell blastocyst injection and generation of chimera and heterozygous (GT/ϩ) mice were performed at the University of Utah Transgenic Gene-targeting Mouse Core Facility. The rd8 mutation (48) was removed by crossing our Cetn3 GT/ϩ animals with C57BL/J WT mice. We mated Cetn3 GT/GT with flippase (Flp) mice (C57BL/6 background) to generate animals with a floxed allele (Cetn3 fl/ϩ ) (Fig. 1A). Exon 3 was deleted by crossing Cetn3 fl/fl with CMV-Cre mice (C57BL/6 background) to generate Cetn3 ϩ/Ϫ and Cetn3 Ϫ/Ϫ mice. Mice were maintained under 12-h cyclic dark/light conditions. ES cell and mouse tail genomic DNA was extracted using a standard protocol (49). Routine PCR genotyping was performed using genomic DNA prepared with HotSHOT (50) as the template, along with genotyping primers and other primers (see below and listed in Table 2).

Electroretinography
ERG was performed on 8-month-old Cetn3 GT/GT , Cetn3 GT/ϩ , and WT controls and on 1-and 3-month-old Cetn2 Ϫ/Ϫ ; Cetn3 GT/GT , Cetn2 ϩ/Ϫ ;Cetn3 GT/GT , and Cetn3 GT/GT animals (n ϭ 5/group) using a UTAS E-3000 universal electrophysiological system (LKC Technologies) as described (23). Briefly, mice were dark-adapted overnight, anesthetized by intraperitoneal injection of ketamine (100 g/g body weight) and xylazine (10 g/g body weight) in 0.1 M PBS, and positioned on a recording platform with body temperature maintained at 37 Ϯ 0.5°C. After pupils were dilated with 1% tropicamide solution (Bausch & Lomb Inc., Tampa, FL), ERG responses were recorded from 5 mice of each genotype/time point. For scotopic ERG, mice were tested at intensities ranging from Ϫ1.63 log cd s/m 2 to 2.38 log cd s/m 2 . For photopic ERG, a rod-saturating background light of 1.3979 log cd s/m 2 was applied for 20 min before and during recording at Ϫ0.01 log cd s/m 2 to 1.86 log cd s/m 2 . Bacitracin ophthalmic ointment (Perrigo, Minneapolis, MN) was routinely applied to the eye to prevent infection after ERG testing, and animals were kept on a heating pad until fully recovered before being returned to cages. Peak amplitudes for both a-and b-waves were used for analysis using a one-way ANOVA test.

Measurement of ONL thickness and OS length
Average ONL thickness and OS length were measured based on DAPI and anti-GC1 antibody fluorescence as described (51). Eyes from WT or knockout mice were removed, marked on the nasal side for orientation, cut into cryosections, and labeled with DAPI and anti-GC1 mAb. The outer nuclear layer and outer segment layers were defined by DAPI and GC1, respectively. Three measurements of the outer nuclear layer and outer segment layers were taken every 200 m from the optic nerve and averaged. The optic nerve was defined as 0 m.

Confocal immunolocalization
The incubation of cryosections with antibodies and confocal imaging was performed as described (52) with minor modification. For photoreceptor OS proteins, eyecups were fixed by immersion in ice-cold 4% paraformaldehyde, cryoprotected in 30% sucrose, and embedded in OCT compound. Sections (12-m thick) were cut using a Microm cryostat and mounted on charged Superfrost Plus slides (Fisher). For antibodies against ciliary markers CETN3, CETN1, Ac-tubulin, and CEP290 and against SPATA7, RPGR, and RPGRIP, eyecups were embedded directly in OCT compound and cut into 12-m sections. Sections were fixed in Ϫ20°C methanol for 20 min or 4% paraformaldehyde for 5 min before immunolabeling. For RPGR and RPGRIP, the sections were further treated with 0.5% SDS for 5 min after paraformaldehyde fixation. Sections were washed in 0.1 M PBS, blocked using 10% normal goat serum or 2% BSA and 0.3% Triton X-100 in PBS, and incubated with primary antibodies at 4°C overnight. After the PBS washes, signals were detected using Cy3-or Alexa 488 -conjugated goat anti-rabbit/mouse and/or donkey anti-goat/rabbit/mouse secondary antibody (Jackson ImmunoResearch) and contrasted with 1 l/ml DAPI (Invitrogen). The primary antibodies and their sources, including references and dilutions, were: preab-  (64). Commercial primary antibodies used included rabbit polyclonal anti-ML-and S-opsins (Chemicon, 1: 1500); PDE6 (MOE, Cytosignal, 1:1500); mouse monoclonal cyclic nucleotide-gated channel ␣1 and ␣3 (CNGA1/A3) (1: 1000, NeuroMab, UC Davis); and acetylated ␣-tubulin (T5451, Sigma, 1:1000). All of these were validated by Western blotting and/or immunostaining by the manufacturers and widely used in the literature. Images were captured using a Zeiss LSM-800 confocal microscope, with some images adjusted for brightness and contrast using Adobe Photoshop CS3.

Measurement of CC length and CETN1 signal intensity
Confocal RGB (red, green, and blue) images were split into 8-bit single-channel images using ImageJ 1.52a. For measuring the ciliary length defined by CEP290 and CETN1 co-labeling, CC were outlined with an integrated Find Edges plugin (Process-Find Edges) and then thresholded (Image-Adjust-Threshold) to generate binary images with background removed but the outlines of the majority of the CC kept. The same settings (default mode; low threshold between 10 and 40, high threshold 255) were applied to both the control and the experimental groups. Individual long axes (curved or straight) of the CC were labeled using a Freehand line-drawing tool; ROI manager (Analyze-Tools-ROI Manager) was added, and the length was measured. For analyzing CC CETN1 fluorescence intensity, green channel images were first treated with a median mode filter (radius, 5 pixels) to reduce the background and then thresholded (default mode; low threshold 20, high threshold 255) to generate the binary images. Individual ciliary regions were outlined using the Freehand line tool and added to the ROI Manager. The integrated intensity of ciliary region was measured and exported to Microsoft Excel.

Electron microscopy
2.5-Month-old Cetn3 GT/GT and Cetn2 Ϫ/Ϫ ;Cetn3 GT/GT retinas (n ϭ 3/group) were immersion-fixed for 2 h in fixative (2% glutaraldehyde and 1% paraformaldehyde in 0.1 M sodium cacodylate buffer, pH 7.4) at 4°C, postfixed for 1 h in 1% osmium tetroxide, and stained en bloc with uranyl acetate. The washed specimens were dehydrated through an ascending series of methanol, dried in propylene oxide, and infiltrated overnight with a resin/propylene oxide (1:1) mixture followed by 100% Epon resin for 2 days. Specimens were embedded in plastic, and the plastic was cured by incubation in a 60°C oven for 2 days. Blocks were trimmed, and 1-m-thick sections were cut to orient photoreceptors near the optic nerve. Ultrathin sections at 60 nm were cut, placed onto slot grids with carbon-coated Formvar film (EMS, Hatfield, PA), post-stained with uranyl acetate followed by lead citrate, and finally examined using a JOEL electron microscope at 75 kV. CC length, defined as the distance between the ciliary pocket base and the first ciliary membrane evagination, was measured. The measurement was repeated by a person blinded to animal genotypes to verify the results.

Statistics
Data are presented as mean Ϯ S.D., where n represents the number of mice (ERG and retina measurement) or number of photoreceptors (CC staining analysis). Statistical comparisons (significance level set at p Ͻ 0.05) were performed using oneway ANOVA for all experimental data.