Deletion of CEP164 in mouse photoreceptors post-ciliogenesis interrupts ciliary intraflagellar transport (IFT)

Centrosomal protein of 164 kDa (CEP164) is located at distal appendages of primary cilia and is necessary for basal body (BB) docking to the apical membrane. To investigate the function of photoreceptor CEP164 before and after BB docking, we deleted CEP164 during retina embryonic development (Six3Cre), in postnatal rod photoreceptors (iCre75) and in mature retina using tamoxifen induction (Prom1-ETCre). BBs dock to the cell cortex during postnatal day 6 (P6) to extend a connecting cilium (CC) and an axoneme. P6 retina-specific knockouts (retCep164-/-) are unable to dock BBs, thereby preventing formation of CC or outer segments (OSs). In rod-specific knockouts (rodCep164-/-), Cre expression starts after P7 and CC/OS form. P16 rodCep164-/- rods have nearly normal OS lengths, and maintain OS attachment through P21 despite loss of CEP164. Intraflagellar transport components (IFT88, IFT57 and IFT140) were reduced at P16 rodCep164-/- BBs and CC tips and nearly absent at P21, indicating impaired intraflagellar transport. Nascent OS discs, labeled with a fluorescent dye on P14 and P18 and harvested on P19, showed continued rodCep164-/- disc morphogenesis but absence of P14 discs mid-distally, indicating OS instability. Tamoxifen induction with PROM1ETCre;Cep164F/F (tamCep164-/-) adult mice affected maintenance of both rod and cone OSs. The results suggest that CEP164 is key towards recruitment and stabilization of IFT-B particles at the BB/CC. IFT impairment may be the main driver of ciliary malfunction observed with hypomorphic CEP164 mutations.

Knockdown of CEP164 in zebrafish resulted in syndromic ciliopathy with ventral body axis curvature, cell death, abnormal heart looping, pronephric tubule cysts, hydrocephalus, and retinal dysplasia [13,16]. Homozygous Cep164 mouse knockouts were embryonically lethal [17]. In a conditional knockout mouse model that lacks CEP164 in multiciliated tissues and testes, a profound loss of airway, ependymal, and oviduct multicilia resulted, and the mutant mouse developed hydrocephalus and male infertility [17]. Using tracheal cell cultures of this mouse model, CEP164 was shown to regulate small vesicle recruitment, ciliary vesicle formation and BB docking [17].
Photoreceptor outer segments (OSs) are considered modified primary cilia, dedicated to light reception and phototransduction [18][19][20][21]. As CEP164 forms a complex with multiple ciliary proteins [22], and there is variance in patient phenotype with CEP164 disruption, additional functions may be revealed depending on depletion before or after ciliogenesis. We generated CEP164 conditional knockouts in retina with three different Cre drivers that produce CEP164 truncation during retina embryonic development ( ret Cep164 -/-) pre-ciliogenesis, in rod photoreceptors ( rod Cep164 -/-) post-ciliogenesis, and in adult retina using tamoxifen induction ( tam Cep164 -/-) (S2A Fig). ret Cep164 -/mice were unable to elaborate photoreceptor connecting cilia (CC) and OSs, whereas rod Cep164 -/mice do form photoreceptor CC and OS as Cre expression starts post-ciliogenesis. Depletion of CEP164 post-ciliogenesis results in profound loss of intraflagellar transport (IFT) particles, leading to impairment of IFT and OS degeneration. We propose that, in ciliopathies caused by missense mutations (e.g., X1460WextX57 associated with Leber congenital amaurosis (LCA), impairment of IFT provides a mechanism for ciliary dysfunction and disease.

Generation of retina-specific Cep164 knockout mice
The murine Cep164 gene (30 exons, one noncoding exon 1) produces multiple splice variants lacking internal exons. The main variant encodes a protein of 1333 amino acids (150 kDa) carrying an N-terminal WW domain and three coiled-coil domains. To generate a retina-specific knockout mouse at various stages of development (S2A Fig), Cep164 F/F mice [17] were bred with transgenic Six3Cre mice ( ret Cep164 -/-) [23], iCre75 mice ( rod Cep164 -/-) [24], or Prom1 (C-L) (Prom1-ETCre) ( tam Cep164 -/-) mice [25] to produce null alleles (S2C and S2D Fig). Six3 (sine oculis-related homeobox 3) is a transcription factor expressed in retina, RPE and brain during embryonic development [23]. Cre in iCre75 is under control of the rhodopsin promoter that expresses in the second postnatal week, allowing CC and OS to develop. Tamoxifen induction allows depletion of CEP164 at any time during development, including in the adult mouse. LoxP sites placed in introns 3 and 4 specify deletion of exon 4 (S2C Fig), truncating CEP164 after exon 3 (after residue Pro 64 inside the WW domain) as exon 5 is out-of-frame. Truncation after exon 3 is predicted to eliminate all splice variants and the WW domain interacting with TTBK2. Genotyping with tail and retina DNA as templates and primers flanking exon 4 verified knockout of the Cep164 gene (S3 Fig). Immunohistochemistry in P21 wildtype (WT) retina with anti-CEP164 antibody showed that CEP164 is a photoreceptor inner segment (IS) protein (S1A

Cep164 -/photoreceptors lack connecting cilia and outer segments
Six3Cre expression begins at embryonic day 9 (E9) and the floxed gene of interest is deleted subsequently. In P6 Cep164 F/+ ;Egfp-Cetn2 controls, BBs docked to the cell cortex as evidenced by presence of CC distal to BBs (Fig 1A). CEP164 appears firmly attached to BBs (Fig 1D, arrows) and short CC are formed (Fig 1D, arrowheads). Few ret Cep164 -/-BBs carry CEP164 and CC formation declines in central ret Cep164 -/photoreceptors (Fig 1B and 1E), presumably because BBs are unable to dock to the IS apical membrane. Similarly, in primary cultures of tracheal multiciliated cells from Cep164 F/F ;FoxJ1-Cre mice, up to 83% of BBs were undocked cryosections on Egfp-Cetn2 + background probed with anti-CEP164 antibody (red). In ret Cep164 -/cryosections, CEP164-BB association is much reduced, but not in retina periphery where degeneration is delayed. D-F, enlargements of A-C as indicated. CEP164 associates with the BB (arrows) in control sections (D), is absent in the knockout (B'), but detectable at some BB in the retina periphery (F). Arrowheads identify CC. G-I, P16 cryosections probed with anti-CEP164 antibody. Note reduced thickness of mutant ONL. J-L, enlargements as indicated in G-I. At P16, control CC (G, J) are normal length and CEP164 is firmly associated with the BB (arrows in J). [17]. As Six3Cre expression generates a central-to-peripheral gradient of gene product depletion [26], knockout is delayed in peripheral ret Cep164 -/retina (Fig 1C), and consequently some photoreceptors develop short CC (about 1/3) (Fig 1F, arrowhead). Mother centrioles with bound CEP164 (Fig 1J, arrow) extend fully developed CC in P16 control photoreceptors (Fig 1G and 1J, arrowhead), whereas in central knockout retina, BBs disintegrate and photoreceptors degenerate (Fig 1H and 1K). Few BBs with CC survive in the retina periphery (Fig 1I  and 1L, arrow in L). Quantitative evaluation reveals that 93.3% control BBs, but only 3.9% knockout BBs, were labeled by antibody directed against CEP164 in P6 central retina (n = 3, Fig 1M). Labeling increased in the peripheral retina to 36.6%. Absence of 100% labeling in the control is attributed to the imaging plane since CC formation was robust. Thickness of ret Cep164 -/outer nuclear layer (ONL) appears to be stable at P6 and P10, but declines by P16 (n = 3, Fig 1N). ret

Cep164 -/photoreceptor function at P16 and P25
Antibodies directed against PDE6 and ARL13B were used to assay for OS elaboration in central versus peripheral retina. PDE6 is a key phototransduction enzyme [27] and ARL13B is a small GTPase functioning as a guanine nucleotide exchange factor (GEF) of ARL3 [28]. In cryosections of Cep164 +/F ;Egfp-Cetn2 control retina, PDE6 (MOE antibody) and ARL13B localize in the photoreceptor OS as expected (Fig 2A and 2D). In central ret Cep164 -/sections, PDE6 (Fig 2B) and ARL13B (Fig 2E) mislocalize to the IS as OSs fail to form. In peripheral retina (Fig 2C), PDE6 localizes to rudimentary OSs while ARL13B remains mostly confined to IS (Fig 2F). OS formation in the retina periphery encouraged functional analysis by pan-retina electroetinography (ERG) allowing assessment of photoreceptor function in dark-and lightadapted retina [29]. In response to light, photoreceptors hyperpolarize, causing a change in electric activity that can be detected by ERG, a diagnostic test called employing light flashes of various intensity [29]. Scotopic a-wave (rod function) and photopic b-wave amplitudes (cone and bipolar cell function) were assessed with flashes of 1.4 log cd s�m -2 . Scotopic and photopic responses were almost extinguished at P16 and P25 indicating loss of rod and cone function (n = 3-6, Fig 2G and 2H). High flash intensities under scotopic conditions elicit weak a-wave responses (Fig 2I) and under photopic conditions an even stronger response (Fig 2J) consistent with minor OS formation and photoreceptor survival at the retina periphery.

Cep164 rod knockout
Previous reports dealt with CEP164's role in early steps of ciliogenesis, such as preciliary vesicle recruitment and BB docking, but its role in ciliary maintenance is unknown. To determine whether CEP164 is required once CC are established, we examined deficits in a rod-specific knockout of CEP164 driven by iCre75 under control of the rhodopsin promoter (S2A Fig).
One advantage of a rod-specific knockout is that Cre expression does not occur before P7, thus permitting ciliogenesis and OS formation. We asked, whether CEP164 is necessary for continued rod viability and, if so, how fast rods would degenerate after CEP164 deletion. Genotyping of rod Cep164 +/control and Cep164 F/F ;iCre75 (abbreviated rod Cep164 -/-) retina DNA confirmed that Cep164 exon 3 excision was complete at P16 (S3G Fig, lane 5).

Loss of CEP164 at rod Cep164 -/-BB
Immunohistochemistry at P12, P15, P21 and P30 revealed that CEP164 was firmly attached to the BB distal appendages in Cep164 +/F ;Egfp-Cetn2 cryosections (Fig 4, rows A and B). At P12, rod Cep164 -/-BBs reveal normal appearance with CEP164 firmly attached. At P15 most BBs have lost most CEP164 and at P21, CEP164 was undetectable at BBs (Fig 4, rows C and D). The P30 knockout ONL revealed one layer of nuclei; a single BB/CC structure is likely that of a surviving cone (Fig 4D, P30

Cep164 -/photoreceptor function
Scotopic (rod function) and photopic (cone function) pan-retina ERGs were performed at a constant 1.4 log cd s/m 2 intensity and the traces averaged (N = 4-14) (Fig 5A and 5D). In scotopic ERG with control animals, rod a-waves continue to increase from P15 (120 μV) to P30 (300 μV) while rod Cep164 -/rod a-waves hover at 50 μV at P16 and P21 (Fig 5A and 5B). At P30, the rod Cep164 -/rod a-wave is nearly extinguished, consistent with absence of rods ( Fig  5B). The photopic rod Cep164 -/cone b-wave is identical to control at P15, reduced at P21 and extinguished at P30, indicating cone degeneration (Fig 5D and 5E). This is consistent with previous findings where deletion of rod-specific genes resulted in secondary loss of cones due to diminished rod-derived cone viability factor (RdCVF) [30]. Scotopic ERG as a function of light intensity showed diminished rod a-and b-wave amplitudes as early as P15 beginning at 1.4 log cd s/m 2 (Fig 5B and 5C) suggesting rod degeneration has begun, although not statistically significant at that age. Photopic cone b-wave amplitudes of control and rod knockout mice were identical at P15 but those of knockouts were nearly extinguished by P30 (Fig 5E  and 5F). Assuming that cone bipolar cells are unaffected in rod Cep164 -/retinas, the severely attenuated photopic cone b-wave derives from loss of cone function. Cone degeneration in the rod knockout was verified using anti-ML-opsin antibody (Fig 5G and 5H). P21 rod Cep164 -/cones are relatively stable, but disintegrate by P30.

Loss of IFT88, IFT57 and IFT140 at BB and CC
In primary cilia (RPE1 cells), CEP164 recruits NPHP1 and IFT88 to the mother centriole and, upon deletion of CEP164, the levels of BB-associated IFT88 strongly decrease [31]. In contrast to primary cilia, CEP164 is dispensable for the recruitment of IFT components to multicilia [17]. We explored whether IFT88, IFT57 and IFT140 levels vary in P16 rod Cep164 -/retinas (Fig 6A-6E) before onset of photoreceptor cell and OS degeneration, and at P21 when OSs disintegrate and ONL thickness is reduced (see Fig 3). P16 and P21 control rod Cep164 +/photoreceptors reveal the familiar pattern of IFT proteins positioned at the BB and CC distal tips (Fig 6A-6E), upper panels) [32][33][34]. By contrast, P16 rod Cep164 -/photoreceptors reveal reduced levels of IFT proteins at BB and CC, although not statistically significant (n = 3, Fig  6A and   Cep164 -/cryosections probed with CEP164 antibody. Row D, enlargements as indicated by dashed squares in row C. Note CEP164 is attached to rod Cep164 -/-BB at P12, but lost at P15 and P21. BB and CC appear perfectly normal at P15 and P21 in the absence of CEP164. At P30, rods are degenerated. Scale bar, 10 μm in rows A and C, 5 μm in rows B and D. https://doi.org/10.1371/journal.pgen.1010154.g004 . Note knockout a-and b-wave amplitudes were diminished at all ages. B, statistical evaluation of amplitudes of control and knockout CEP164 scotopic ERG a-waves at P15, P21 and P30 as measured in A. At P21 and P30 � p<0.05 t-test C, F, scotopic a-wave (C) and photopic b-wave amplitudes (F) at P16, P21 and P30 as a function of light intensity. Knockout scotopic a-and b-waves are diminished, but never extinguished. � p<0.05 at P21 and^p<0.05 at P30. D, control (blue) and rod Cep164 -/mice (red) were exposed under photopic conditions to flashes of 1.4 log cd s/m 2 . Note, rod Cep164 -/photopic b-waves were absent at P30. E, statistical evaluation of photopic b-wave amplitudes at 1.4 log cd s/m 2 as measured in B. G, H, immunohistochemistry with anti-ML-opsin of control (G) and rod Cep164 -/central cryosections (H) at P16, P21 and P30. Cones on the rod Cep164 -/background are stable until P21 but are degenerated at P30. knockouts (n = 3, Fig 6C and 6D, lower panels, Fig 6F). IFT88, IFT57 and IFT140 fluorescence signal intensities at distal CC in rod Cep164 -/retina are reduced at P16, although not statistically significant (S5 Fig), and to 30% or less by P21 (Fig 6F). SPATA7 (spermatogenesis associated protein 7) [35] and CEP290 [36,37] are unaffected by deletion of CEP164 (Fig 7A, 7B and 7F) Cep164 -/-(lower panels) cryosections at P16 probed with anti-IFT88 (A) and anti-IFT57 (B) antibodies. C-E, immunohistochemistry of P21 rod Cep164 +/-(upper panels) and rod Cep164 -/-(lower panels) cryosections probed with anti-IFT88 (C), anti-IFT57 (D) and anti IFT140 (E). Insets to the right of main panels show enlargements of individual representative BB/CC structures as indicated by hatched rectangles (left panels) or antibody alone (red). F, red channel pixel intensity of IFT88, IFT57 and IFT140 signal at CC distal end at P21 using 3 animals each. The IFT57 signal was not significant, likely due to higher variance in the control. suggesting normal CC elaboration. NPHP1 remains unaffected by post-ciliogenesis deletion of CEP164 (Fig 7C and 7F). FBF1 remains at the BB in the rod Cep164 -/-, suggesting intact DAPs (Fig 7D). These results suggest that CEP164 recruits IFT88, IFT140 and IFT57 to photoreceptor BB/CC, as observed in primary cilia [31]. IFT88 and IFT57 protein levels are reduced in the rod-Cep164 -/-, negatively affecting IFT and OS axoneme stability as evidenced by shortening of rod-Cep164 -/rod OSs at P16 and P21 (Fig 3B and 3D) and absence of P14 discs in the dye injection experiment (see below, Fig 8C).
https://doi.org/10.1371/journal.pgen.1010154.g008 previously shown to be released from the centrosome to move along microtubules in epithelial cells providing a potential explanation [46].

Disc formation in rod Cep164 -/photoreceptors
All~800 discs of the mouse OS are renewed every ten days and roughly 80 discs are synthesized every day, moving towards the retina pigmented epithelium (RPE) to be phagocytosed [47]. To identify mechanisms that cause OS degeneration more precisely, we developed a fluorescent dye assay to monitor the formation of closed discs during disc morphogenesis and their fate. As Cre expression varies in each photoreceptor, to some extent, and dye sequesters into new discs in OS per photoreceptor, dissociated photoreceptors were examined independently. CF-568-hydrazide, a water-soluble, membrane impermeable, aldehyde-fixable redfluorescent dye [48], was injected intravitreally on P14 (day 1) and P18 (day 4) and eyes were harvested on P19 (day 5). Following retina dissociation, rod Cep164 +/control photoreceptors show two bands containing red fluorescent discs along the OS (Figs 8A and 3B) indicating that CF-568 sequestered into the lumen during enclosure of a nascent disc. The two bands are separated by about 10 μm or~300 discs. Nascent day 1 discs (P18) are located immediately adjacent to the CC (green, identified by EGFP-CETN2), whereas older day 4 discs (P14) are displaced distally towards the retinal pigment epithelium (RPE), as shown by Richard Young [49,50]. Disc morphogenesis occurs as late as day 4 (P18) in rod Cep164 -/-OSs, albeit slower than controls as newly formed discs appear closer to the distal CC (n = >100, Fig 8C and 8D). Relative to control OSs, day 4 rod Cep164 -/bands are very faint or absent, i.e., only~5% of KO rods display P14 bands (n = >100, Fig 8E).
Labeled by anti-PDE6 antibody (MOE), tam Cep164 -/rod OSs degenerated slowly and were approximately half-length at 3wPI (Fig 9A, right panel) relative to control OSs (Fig 9A, left and middle panels). tam Cep164 -/rod OSs probed with anti-ARL13B (Fig 9B, right panel) revealed that ARL13B is unable to reach the OS in the absence of CEP164, confirming the interaction of these two proteins [22]. tam Cep164 -/cone OSs identified by anti-ML-opsin were shorter at 3wPTI (Fig 9C, right panel) compared to controls (Fig 9C, left and middle panels) and ML-opsin mislocalized in part to the cone IS. At 4wPTI, mutant cone OSs were largely degenerated (Fig 9D, right panel). About 60% of tam Cep164 -/rods analyzed by confocal microscopy at 2.5wPTI had BBs associated with CEP164, while the rest were CEP164-free (n = 3, Fig 9E and 9F) indicating a 40% induced knockout efficiency. Anterograde IFT57 and IFT88 showed normal distributions at control BB and CC tips (Fig 10A and 10B, upper panels) (Fig 6A and 6B). As expected, IFT57 and IFT88 were absent at the tam Cep164 -/-CC tip and reduced at BB (Fig 10A and 10B, lower panels), similar to the CEP164 rod knockout at P16. NPHP1 localization at CC was unaffected (Fig 10C).

Photoreceptor CEP164
Average tam Cep164 -/scotopic a-wave amplitudes, not significantly reduced at 2wPTI, were diminished by 3wPTI (n = 5-7, Fig 11A) indicating loss of rod function. Statistical evaluation of scotopic a-waves at 2-4wPTI at 1.4 log cd s/m 2 confirmed diminished a-wave amplitudes (50% and 45%, respectively) (n = 3-7, Fig 11C). A-wave amplitudes did not significantly change at a higher light intensity of 2.4 log cd s/m 2 for 2wPTI (n = 3-7, S7A and S7B Fig). Photopic tam Cep164 -/b-waves persistently diminished from 3wPTI down to 40% of control at 4wPTI (Fig 11B and 11D). A-and b-wave amplitudes were not reduced further at 6wPTI and 9wPTI (S7C and S7D Fig). Pan-retina ERG as a function of light intensity (-4.5 to 2.4 log cd s/ m 2 ) did not reveal significant differences of scotopic/photopic a-and b-wave amplitudes at 2wPTI (S8A-S8C Fig). At 3wPTI, scotopic ERG showed diminished a-and b-wave amplitudes beginning at 0.6 log cd s/m 2 . Photopic b-wave amplitudes of control and tam Cep164 -/mice were statistically identical at 2wPTI, but decreased by 3wPTI (S8C Fig).

Photoreceptor CEP164
CEP164 has no effect on pre-ciliogenesis photoreceptor differentiation or ONL lamination (Fig 1N), but prevents OS formation (Fig 2B and 2E), essentially confirming the well-documented role of CEP164 in cultured cells. Loss of CEP164 early leads to rapid rod and cone degeneration centrally with an LCA-like phenotype (Fig 2G-2J).
Because CEP164 disruption results in variable patient phenotypes, we reasoned that CEP164 may have additional functions to be revealed during post-ciliogenesis deletion of CEP164. Two drivers express Cre after ciliogenesis (iCre75 and tamoxifen inducible Prom1-CreER T2 ) when CEP164 is bound to distal appendages and the BB is docked to the apical membrane. Postnatal depletion of CEP164 in rods with iCre75 allows the BB to dock to the cell cortex. Control photoreceptor CC extend axonemes around P6, and OSs are mature at P21. P16 rod Cep164 -/plastic sections do not reveal major morphological changes compared to controls, but the ONL is reduced at P21 (S4 Fig), confirmed by statistical analysis comparing control and rod Cep164 -/-ONL thickness (Fig 3E). At P30, only one nuclear row is present in the knockout (Fig 3B and 3D). PDE6 mislocalization in the IS was not obvious in the P21 rod-Cep164 -/photoreceptors when OSs are disintegrating (Fig 3B), suggesting continuous delivery of PDE6 by its chaperone PDE6D to OSs as OS discs continue to be formed. By contrast, ARL13B is absent in rod Cep164 -/-OSs consistent with the observation that ARL13B interacts with CEP164 (Fig 3D) [22]. The results suggest that the BB-CC-axoneme structure at the OS base is relatively stable even as the OS degenerates, in contrast to CEP164's ciliogenesis role where CEP164 is required for recruitment and physical attachment of small preciliary vesicles to the mother centriole and membrane docking. While P16 control and rod Cep164 -/retina morphologies are comparable (S4A Fig), the scotopic a-wave already is reduced at P15 indicating rod degeneration has already begun (Fig 5B), although not statistically significant at this age. In P15 and P21 rod Cep164 -/photoreceptors, most BB have lost CEP164 (Fig 4B and 4B) but P16 OSs still possess BB-CC structures (Fig 3B).
A major finding is that IFT88, IFT57 and IFT140 are depleted gradually after deletion of CEP164 at P16-P21 (Figs 6 and S5), implicating IFT malfunction as the cause of retina degeneration in rod Cep164 -/mice. Similar results were seen in tamoxifen-induction experiments at 2.5WPTI (Fig 10A and 10B). Changes in localization and expression levels of PCNT and ninein (S6 Fig) suggest that other PCM and subdistal DAPs may be affected by CEP164 deletion. IFT is integral for assembly, maintenance, and function of motile and primary cilia [53,54], relying on molecular motors and IFT particles to move cargo along microtubules in anterograde (heterotrimeric kinesin-2) or retrograde (dynein-2) direction. IFT88 (Tg737) and IFT57 are required for ciliary and Chlamydomonas flagella assembly [55]. IFT57 stabilizes the IFT-B complex in Chlamydomonas flagella [56] and interacts with IFT20 and the kinesin-2 subunit, KIF3B [57]. Germline deletions of IFT88 are embryonically lethal [58,59]. In mouse photoreceptors, a hypomorphic mutation in the IFT88 gene exhibited abnormal OS development and retinal degeneration demonstrating that IFT88 is important for OS maintenance [32]. In conditional retina IFT88 knockouts ( ret Ift88 -/-), BBs failed to extend CC [60]. Loss of IFT88 (Tg737) and IFT57 (Hippi) during CEP164 deletion confirms CEP164's involvement in recruiting these proteins to the photoreceptor BB, first observed in primary cilia [31]. Another DAP protein involved in recruiting IFT-B proteins (IFT88 and IFT52) is C2CD3 (OFD14) which localizes to the distal ends of both mother and daughter centrioles. C2CD3 was shown to recruit IFT88 and IFT52 to the mother centriole [4]. As C2CD3 is required for the recruitment of DAP proteins including CEP164, the effect of recruiting IFT-B proteins may be indirect. Loss of C2CD3 results in shortened centrioles without appendages and failure of CP110 removal from the ciliary mother centriole, a critical step in initiating ciliogenesis. This provides a potential mechanism for the later onset of retinal degeneration in patients with hypomorphic CEP164 mutations.
Instability of the OS distal tip with continued disc morphogenesis proximally was also seen in the CF-568 dye assay that allows monitoring of disc morphogenesis, enclosure and displacement in mouse rods (Fig 8). Intravitreally-injected CF-568-hydrazide [48] is sealed into the disc lumen during the disc enclosure process. The number of labeled discs, limited by available dye, is fewer than 80, the number of discs generated per day. Although disc morphogenesis continued as late as P18, the rate of disc renewal was impaired (Fig 8E). We propose that instability at the rod Cep164 -/-OS distal tip may be caused by axonemal disintegration as IFT-B proteins dissociate from BB thereby impairing IFT (Figs 6 and 8C).
IFT cargo in rods is controversial. Current models suggest that rhodopsin and other integral membrane proteins traffic by anterograde IFT, mediated by kinesin-2, along the CC for deposition into nascent discs [61][62][63][64]. In our hands iCre75-based conditional knockout of KIF3A, the kinesin-2 obligatory subunit, did not prevent trafficking of rhodopsin and other proteins to the OS [65]. Conditional knockout of KIF3A with Six3Cre prevented CC and axoneme formation, although BB docked to the cell membrane correctly [60]. Conditional knockout of KIF3A and IFT88 in the adult mouse by tamoxifen injection demonstrated that photoreceptor axonemes disintegrate slowly post-induction, starting distally, but rhodopsin and cone pigments traffic normally for more than 2 weeks, a time interval during which the OS is completely renewed. Thus, visual pigments likely transport to the photoreceptor OS despite KIF3A, IFT88 removals and IFT impairment. Alternatively, kinesin-2-mediated anterograde IFT appears to be responsible for photoreceptor transition zone and axoneme formation [60].
CEP164 forms a complex with PDEδ (a prenylated protein chaperone), INPP5E (a farnesylated inositol polyphosphate 5 phosphatase) and ARL13B (GEF of ARL3) to traffic prenylated INPP5E to primary cilia in cell culture [22]. However, INPP5E of WT mouse photoreceptors localizes to the IS, is excluded from the OS, and is apparently independent of PDEδ [66]. Absence of INPP5E in the OS would seem to nullify experiments to show that ARL13B, PDEδ or CEP164 act as chaperones for INPP5E. Connecting cilia and short rudimentary OSs form in ret Inpp5e -/photoreceptors, but without axonemal structure and discs [66]. ARL13B localizes to WT OSs, but in ret Arl13b -/- [67] and ret Cep164 -/photoreceptors (Fig 2E), CC and OS fail to form and ARL13B accumulates in the IS. In rod Cep164 -/photoreceptors, OSs form but ARL13B is barely detectable at P16 and absent at P21 (Fig 3D), while in tam Cep164 -/photoreceptors, ARL13B is depleted in OS (Fig 9B, right panel). These observations are consistent with [22] who show that ARL13B and CEP164 interact to enable ARL13B to reach the cilium, but not with [17] where ARL13B accumulated in CEP164-KO multiciliated cells. Thus, the functional network consisting of ARL13B, INPP5E, PDEδ and CEP164 is an arrangement of primary cilia for INPP5E and ARL13B to traffic to cilia, but is conserved only partially in photoreceptors.

Summary
Deletion of CEP164 in photoreceptors before ciliogenesis prevented BB docking as shown in primary cilia, leading to rapid rod/cone degeneration resembling LCA. Deletion of CEP164 after ciliogenesis and BB docking reveals that BB/CC/OS structure lacking CEP164 is initially stable. Post-ciliogenesis deletion of CEP164 causes loss of IFT proteins and impairment of IFT, an essential transport mechanism for OS assembly and maintenance. Impairment of IFT is demonstrated by a fluorescent dye assay that labels newly generated discs. In rod Cep164 -/rods, discs that travel to the mid-distal OS degenerate, while disc morphogenesis continues proximally at a slower rate, consistent with axoneme damage and IFT impairment. We favor a model in which the rod Cep164 -/axoneme is ill-maintained as IFT-B proteins dissociate from the BB, causing OS degeneration as IFT trains fail. Loss of IFT88 and IFT57 during CEP164 deletion suggests that CEP164 recruits anterograde IFT proteins to the BB as proposed by [31], and identifies CEP164 and DAP structure as integral for initiating IFT.

Ethics statement
All mice were handled in accordance with NIH guidelines, and all protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Utah (Protocol 18-11005).

Generation of knockouts by tamoxifen induction
Cep164 F/F mice were bred to Prom1-CreER T2 mice [25] to generate Cep164 F/+ ;Prom1--CreER T2 mice. Expression of CreER T2 is driven by the prominin 1 (Prom1) promoter/ enhancer. Cep164 F/+ ;Prom1-CreER T2 mice were bred to Cep164 F/F mice to generate mice for tamoxifen injection. Tamoxifen was administered via intraperitoneal injection in adult mice (P28-P30 at time of first injection). Tamoxifen was dissolved in corn oil to a stock solution of 20mg/ml. Mice were dosed with 150 mg/kg body weight for 5 consecutive days according to their weight on the first day of injection (i.e., 7.5 μL of 20mg/ml tamoxifen solution per gram weight). ERG and eye collection for confocal immunolocalization were performed 2-9 weeks after the first injection.

Genotyping
Genomic DNA was extracted from fresh tissue by dissolving tail clips or retina DNA from P8-12 day old mice in 150 μL tail lysis buffer at 50-60˚C for 1-2 hour. Digests were centrifuged at 15000 rpm for 5 minutes. Supernatant was added to an equal volume of isopropanol and centrifuged at 15000 rpm for 5 minutes. The DNA pellet was rehydrated in 75 μL H 2 O. Genotyp

Intensity measurements
We used the average pixel intensity of a drawn region of interest (ROI) function in Zen Blue. For IFT markers, a circle ROI was drawn around the distal tip of the CC, as determined by the Cetn2-EGFP signal. Over 30 cells were counted across 3 different confocal images and averaged per mouse. For CC markers, a rectangular ROI was drawn around the CC and average intensity measured. We performed a two-tailed Student's t-Test to compare controls to knockouts. Microsoft Excel function "T.TEST assuming equal variances, two-tailed" was used to calculate the p-value, with statistical significance determined using an alpha value of p < 0.05.

Electroretinography
Scotopic and photopic ERG measurements were performed using P16 and P25 for Six3Cre experiments and P16, P21, or P30 mice for the iCre75 experiments and at 2, 3, 4, 6 or 9 weeks post-tamoxifen induction for the Prom1-CreER T2 experiments. Prior to ERG the mice were dark-adapted overnight and anesthetized with intraperitoneal injection of 1% ketamine/0.1% xylazine at 10 μl/g body weight. The mice were kept warm during ERG by using a temperature-controlled stage. Scotopic and photopic responses were recorded as described [66] using a UTAS BigShot Ganzfeld system (LKC Technologies, Gaithersburg, MD). Scotopic singleflash responses were recorded at stimulus intensities of -4.5 log cd s�m -2 [log candela seconds per square meter] to 2.4 log cd s�m -2 ). Mice were light-adapted under a background light of 1.48 log cd s�m -2 for 5 minutes prior to measuring photopic responses. Photopic single-flash responses of control and knockout were recorded at stimulus intensities of -0.1 log cd s�m -2 to 1.9 log cd s�m -2 .

Statistical analysis
We performed an unbalanced two-factor ANOVA to compare mutant and control mice for their quantified a-and b-wave ERG response across multiple ages or weeks post-tamoxifen induction. Post-hoc multiple comparison was performed using Tukey's honestly significant difference criterion. Statistical significance was determined using an alpha value of p < 0.05. ERG statistics were computed using MATLAB's statistical toolbox "anovan" and "multcompare" functions.
We performed a two-tailed Student's t-Test assuming equal variances to compare ONL thickness for ret Cep164 +/-, ret Cep164 -/-, rod Cep164 +/and rod Cep164 -/-. Retina measurements used for these calculations were determined based on an average of three measurements per retina. Microsoft Excel function "T.TEST assuming equal variances, two-tailed" was used to calculate the p-value, with statistical significance determined using an alpha value of p < 0.05. E9 depicts a progenitor cell in which Six3Cre is expressed. At P0-P1, the centrosome consisting of mother (MC) and daughter (DC) centrioles acquires a ciliary vesicle (CV) distally. From P3-P4, the mother centriole docks to the cortex of the inner segment (IS). During P5-P6, an axoneme emanates from the MC and MC becomes a basal body (green). A-and B-tubules (black) extend at P7-P8, and the transition zone (TZ) becomes established. As plasma membrane evaginates and discs assemble, the rod outer segment (OS) is considered mature at three weeks of age (P21). Adapted from [21], with permission from Elsevier. B, gene structures of human centrosomal protein 164 kDa (CEP164; upper) and mouse Cep164 (lower) consisting of a single WW (yellow) and three coiled-coiled (CC) domains. Positions of mutations associated with human nephronophthisis (NPHP) and retina degeneration are indicated. Asterisk above mouse CEP164 indicates the truncation point of our conditional knockout. C, schematic of Cep164 exons 1-7 and conditional Cep164 allele with loxP recombination sites flanking exon 4. The FRT site originates from flippase removing the gene trap. D, Cre-induced recombination (driven by Six3Cre, iCre75 or Prom1-ETCre) excises exon 4 and generates a null allele. Sigma-Aldrich antibody (C). Pericentrin locates to the pericentriolar material surrounding the BB (yellow); PCNT levels appear slightly reduced in rod Cep164 -/-. Ninein is a subdistal appendage protein associated with centrioles in tissue culture; in rods ninein is very weak at the BB but present in the proximal axoneme of control rods (B, upper panels) or rootlet/microtubule area (C, upper panels) (arrows in B and C enlargements). Panels to the right, enlargements of representative BB/CCs identified by EGFP-CETN2 (green) or individual antibodies (red).